doctoral thesis proposal systems optimization in mobility...
TRANSCRIPT
Doctoral Thesis ProposalSystems Optimization in Mobility Management
Ashutosh DuttaDepartment of Electrical Engineering
Columbia [email protected]
March 14, 2008
i
Abstract
Mobile networking technology seeks to preserve the user experience as the user moves andthereby crosses from one network into another. A mobility event is the result of one network con-nection path being replaced by another via the rebinding of common system properties that mayhappen at any one of several protocol layers of the mobile node. The rebinding is a sequential pro-cess and each process takes a finite amount of time. This overall process generates a period of timein which network service is degraded by transient data loss and increased end-to-end delay. Applica-tion specific and protocol specific ad hoc solutions are available for mitigating the service disruption.However, formal techniques to characterize this problem and to develop optimization methodologiesfor these processes have not been studied as part of a systematic framework for mobility solutions.This dissertation develops a systematic and formalized systems model that analyzes the basic opera-tions associated with a mobility event and characterizes several systems optimization techniques forthese processes.
The proposed formal mobility systems model represents these basic operations in the form ofa discrete event dynamic system. I analyze this general mobility systems framework and developseveral methodologies that can model systems optimization techniques. In particular, I developmethodologies to model the optimization for many of the basic operations, such as discovery, con-figuration, binding update, and media redirection. Then, I apply these methodologies to prototypemobility systems that can support subnet, domain and inter-technology roaming for real-time andstreaming applications in the wireless Internet. Some of these methodologies include proactive net-work and resource discovery, pre-authentication, pre-configuration, reduction of binding delay, andminimizing the effect of media redirection delay by means of dynamic buffering and multicasting.
I validate the mobility systems model by using it to analytically assess related mobility protocolsfor both interactive and multicast streaming traffic. The model can result in an analytical assessmentof the techniques that can be directly compared to our experimental results under certain resourceconstraint and various networking parameters, such as mobility rate, packet-to-mobility ratio, simul-taneous mobility, distance between the communicating nodes, and network access characteristics. Iperform a comparative analysis of our optimized mobility management schemes with similar net-work layer mobility protocols and other fast-handoff mechanisms.
ii
Contents1 Introduction 1
2 Introduction to Mobility Management for Multimedia 2
3 Motivation 3
4 Systems Analysis of Mobility Event 44.1 Analysis of handoff components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54.2 Effect of handoff across layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5 Modeling Mobility as Discrete Event Dynamic System 9
6 Proposed Mobility Optimization Techniques 136.1 Application layer discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146.2 Network layer assisted pre-authentication . . . . . . . . . . . . . . . . . . . . . . . 156.3 Anchor assisted security association . . . . . . . . . . . . . . . . . . . . . . . . . . . 166.4 Layer 3 Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.4.1 Router assisted duplicate address detection . . . . . . . . . . . . . . . . . . . 186.4.2 Proactive address caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.5 Route optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196.5.1 Maintain direct path by application layer mobility . . . . . . . . . . . . . . . 196.5.2 Interceptor-assisted packet modifier . . . . . . . . . . . . . . . . . . . . . . . 206.5.3 Midcom proxy-assisted route optimization . . . . . . . . . . . . . . . . . . . 20
6.6 Binding update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216.6.1 Hierarchical binding update . . . . . . . . . . . . . . . . . . . . . . . . . . . 216.6.2 Proactive binding update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
6.7 Redirection of in-flight data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226.7.1 Mobility-proxy assisted time-bound data redirection . . . . . . . . . . . . . 226.7.2 Small group multicasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
6.8 Network buffering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246.9 Cross-layer triggers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246.10 Cross layer mobility optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256.11 Systems evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.11.1 Intra technology handoff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266.11.2 Inter technology handoff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
7 Roadmap for Future Work 287.1 Verification of optimal handoff operation using Petri net models . . . . . . . . . . . 287.2 Analysis of optimal proactive handoff . . . . . . . . . . . . . . . . . . . . . . . . . . 287.3 Tradeoff analysis between in-flight packet loss and one-way delay . . . . . . . . . . 287.4 Summary of plan for completion of research . . . . . . . . . . . . . . . . . . . . . . 30
8 Conclusion 30
iii
List of Figures1 Functional components of a sample infrastructure-based mobility . . . . . . . . . . . . . 42 A generalized Petri net model for a mobility event . . . . . . . . . . . . . . . . . . . . . 113 Layered modeling with Petri net . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 Petri net representation of mobility event using Time Net . . . . . . . . . . . . . . . . . 135 Petri net mapping for sequence of events . . . . . . . . . . . . . . . . . . . . . . . . . . 146 Cross layer mobility optimization scenario and results . . . . . . . . . . . . . . . . . . . 267 Resource utilization for sequential, concurrent and proactive operations . . . . . . . . . 298 Effect of network buffering on in-flight packets . . . . . . . . . . . . . . . . . . . . . . 30
List of Tables1 Mapping of basic operations of a mobility event . . . . . . . . . . . . . . . . . . . . . . 52 IP address acquisition delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 Mapping basic handoff operations across layers . . . . . . . . . . . . . . . . . . . . . . 94 Description of Places and Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 Cycle time for Petri net with mobility optimizations . . . . . . . . . . . . . . . . . . . . 136 Network layer assisted layer 2 preauthentication . . . . . . . . . . . . . . . . . . . . . . 167 Effect of duplicate address detection (IPv6) on handoff . . . . . . . . . . . . . . . . . . 188 Inter-domain handoff results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 Experimental validation of route optimization . . . . . . . . . . . . . . . . . . . . . . . 2010 Delay and packet loss during proactive handoff . . . . . . . . . . . . . . . . . . . . . . 2711 Plan for completion of my research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
iv
1 IntroductionUniversal wireless connectivity to communications and information has advanced the world towardsubiquitous computing. In the space of less than thirty years cell phones have become ubiquitous andwireless data access has become common. However, this access has brought with it a variety of tech-nical problems. Radio physics and power constraints, the need to reuse spectrum, economic constraintson facility placement, and service balkanization due to competitive and political factors force us to im-plement wireless systems as cells of limited range. Furthermore, cells may use very different wirelesstechnologies or provide fundamentally different services. We then need handoff mechanisms, oftenat multiple protocol layers, to allow a mobile terminal to move from cell to cell and maintain servicecontinuity.
Mobility can be described as terminal movement resulting in the release of its binding to the currentcell (point of attachment to the network) and establishment of bindings to the new cell being enteredwhile preserving the existing sessions associated with higher level services. The cellular telephonycommunity has long implemented service and technology specific mobility protocols that hand off voicesessions as the user moves from cell to cell. Because voice service quality is highly sensitive to serviceinterruptions, cell-to-cell handoffs in a cellular environment have been highly optimized and are notnoticed by the public. Tripathy et al. [1] summarize some of the handoff technologies associated withcellular mobility.
For IP traffic, the IETF has defined mobility protocols for both IPv4 and IPv6 [2], [3]. However,IP traffic is dramatically more diverse than cellular voice in the range of link layer technologies usedto support IP traffic, the number of economic units supplying IP service as well as the authenticationprotocols and services running above IP. This diversity meant that the IETF could not easily design thehandoff optimization seen in cellular voice into the mobility standards. As a result, unoptimized mobileIP handoffs can take a few seconds to perform a handoff and degrade the desired quality of service inthe process.
IP’s transformation from a service supporting email and file transfer to the base layer for networkconvergence means that the constraints on handover performance are becoming much more stringent.Handovers cannot interrupt real-time services. The mechanisms and design principles needed for build-ing optimized handovers in the context of mobile Internet services are poorly understood and need betteranalysis.
This thesis proposal contributes to a general theory of optimized handover, especially with respectto mobility of Internet-based applications. The contributions fall into four categories:
1. Identification of fundamental properties that are rebound during a mobility event. Analysis ofthese properties provides a systematic framework for describing mobility management and theoperations that are intrinsic to handover.
2. A model of the handover process that allows performance predictions to be made for both anun-optimized handover and for specific optimization methodologies under resource constraints.
3. A series of optimization methodologies that could be applied to link, network, and application lay-ers and preserve the user experience by optimizing a handover. These also include implementationand experimental evaluation of the optimization techniques.
4. Application of a model to represent various optimizations and comparison of the model-basedresults against the experimental data.
This proposal is organized as follows. Section 2 introduces mobility management and discusses therelated mobility protocols that are currently available. Section 3 brings out the issues related to handoffand provides the motivation for the proposed research. I provide a systematic analysis of the mobilityevent and associated handoff components in Section 4. In Section 5, I introduce a formal systems modelthat uses Petri nets to analyze a mobility event and the associated optimization techniques. Section
1
6 describes some key mobility optimization techniques that I have developed and demonstrate theirvalidation by way of experiments and modeling. In Section 7, I provide a roadmap towards completionof my thesis with a list of future work items and a timeline. Finally, Section 8 concludes my proposal.
2 Introduction to Mobility Management for MultimediaA mobility management scheme primarily consists of two components: location management andhandoff management. Location management enables the network to discover the current point of at-tachment of the mobile user so that the new connection can be established when a new multimediacall arrives. Handoff management often known as terminal mobility, allows the network to maintainthe user’s connection binding as the mobile node moves from one attachment point to another in thenetwork. I focus on handoff management in my proposal.
As a mobile goes through a handover process, it is subjected to delay because of the rebinding of itsassociation at several layers of the protocol stack. Delays incurred within each of these layers affect theongoing multimedia application and data traffic within the client. There are several basic operations thatare associated with the re-establishment of the binding process at several layers. These operations canbe affected by several factors, such as access characteristics (e.g., bandwidth, channel characteristics),access mechanism (e.g., CDMA, CSMA/CA, TDMA), re-configuration of identifiers, re-authentication,re-authorization, and rebinding of security associations at all layers.
The handover process can be classified based on access technology and administrative domain.Based on access technology, it can be categorized as horizontal and vertical handover. Horizontal han-dover involves a mobile’s movement between the same type of access networks (e.g., 802.11). Verticalhandover includes movement between different types of wireless access networks of different cell sizes.Supporting terminal handovers across heterogeneous access networks, such as CDMA (Code DivisionMultiple Access), IEEE 802.11, WiMAX (Worldwide Interoperability for Microwave Access) and GPRS(General Packet Radio Service) needs to take into account the diverse quality of service, security andbandwidth parameters associated with each type of access network. Similarly, movement between twodifferent kinds of administrative domains poses a challenge since a mobile needs to re-establish authen-tication and authorization in the new domain.
The mobility management techniques can be implemented at several layers of the protocol stack,such as network layer, transport layer, and application layer. MIPv4 [2] and its several variants, such asMIP-RO (MIP with Route Optimization), and MIP-RR (MIP with Regional Registration) [4], MIP-LR[5], MIPv6 [3], MOBIKE [6] are a few network layer mobility protocols that are defined within the IETF.Cellular IP [7], HAWAII (Handoff Aware Wireless Access Internet Infrastructure) [8], Proxy MIP, IDMP(Intra Domain Mobility Protocol) [9] are a few of the micro-mobility protocols suitable for intra-domainmobility. Intra-domain mobility typically refers to a movement scenario when the mobile’s movementis confined to the administrative domain. MSOCKS [10], TCP-Migrate [11], and SCTP (Stream ControlTransport Protocol) [12] have been designed to take care of mobility at the transport layer. SIP-basedmobility protocol [13] takes care of mobility by means of application layer signaling, such as SIP (Ses-sion Initiation Protocol) [14]. HIP (Host Identity Protocol) [15] defines a new protocol layer betweennetwork layer and transport layer to provide terminal mobility in a way that is transparent to both thenetwork layer and transport layer. Similarly, Varshey et al. [16], [17] and Acharya et al. [18] describemany of the architectural issues associated with mobility support for multicast traffic. Romdhani [19] etal. primarily categorize mobile multicasting into two categories, such as home subscription-based solu-tions and remote subscription-based solutions. Mobile Multicast (MOM) [20], range-based MOM [21]and MIP-based bi-directional tunneling fall under home subscription-based solution, whereas Mobicast[22], Hierarchical SSM [23] and remote subscription-based Mobile IP do not depend upon home agentsand can be listed under remote-subscription-based solutions.
I have provided a survey of the related mobility protocols and the issues in [24], [25]. I have alsoexperimentally verified [26], [27] that these mobility protocols in their current form are not adequate tomeet the performance requirements of real-time traffic and hence, need overall systems optimization.
2
3 MotivationIn order to provide the desirable quality of service for interactive VoIP and streaming traffic, one needsto limit end-to-end delay, eliminate jitter, and reduce packet loss to an acceptable threshold level. Theperformance requirement will vary based on the type of application and its characteristics such as delaytolerance and loss tolerance limit. Different standards organizations have defined the threshold limit forthese metrics. 3GPP TS23.107 [28] defines 4 application classes: conversational, streaming, interactiveand background (e.g., ftp, email) each with different set of end-to-end delay and QoS requirement. Basedon the type of applications (e.g., interactive, streaming, data), these values may vary. For example, forone-way delay, ITU-T G.114 recommends 150 ms as the upper limit for VoIP applications and 400 msas generally unacceptable delay. Similarly, the streaming class has the tolerable packet (SDU) error ratesranging from 0.1 to 0.00001 and the transfer delay of less than 300 ms.
In general, the handoff process contributes to the handoff delay, associated packet loss, jitter, andadversely affects the overall throughput of data traffic because of interruption and retransmission of datacaused due to a change in the network point of attachment. A mobility event contributes to two kinds ofdelays that affect the performance, such as delay between the last packet in the old point of attachmentand first packet in the new point of attachment and one-way-delay of the in-flight packets when bufferingis applied. Similarly, handoff contributes to jitter for the in-flight packets. In-flight packets are definedto be the packets that are in flight during the mobile’s movement. Jitter refers to the variation in inter-packet arrival time of the consecutive packets at the receiver. This is caused due to variation of one-waytraversal delay of the consecutive packets. For real-time communication, if a packet is received after acertain delay threshold, it is also considered lost.
I have verified [26] experimentally that in the absence of any optimization technique, a mobile canbe subjected to a handoff delay between 4 and 17 seconds resulting in transient service interruptionand packet loss. This value could vary depending upon the type of mobility protocol used, types ofhandover, such as vertical or horizontal, and type of access networks, such as 802.11 or CDMA. Thus,it is desirable to conduct a formal analysis of these discrete events that constitute the handoff process,build a system model, develop relevant optimization techniques for these operations, and analyze theresources needed for each of these optimization techniques.
While several mobility protocols have been defined at different layers, to the best of my knowledge,prior to my work there was no formal analysis or system model that can study the basic operationsassociated with mobility events and system optimization techniques.
Following are the key issues that motivate my research:
1. Existing mobility protocols affect the performance of real-time communication because of thesequence of discrete events associated with the handoff event.
2. Existing mobility optimization mechanisms are tightly coupled with specific mobility manage-ment protocols and do not provide a generalized approach to optimization. For example, it isimpossible to apply mobility optimization mechanisms designed for Mobile IPv4 [2] or MobileIPv6 [3] to MOBIKE [6]. Thus, it is desirable to develop a set of formal mobility optimizationmethodologies with specific design criteria that can work with any mobility management protocoland access technologies.
3. Available mobility management techniques do not provide any systematic framework to formalizedifferent states and transition processes involved with a mobility event. Thus, it becomes difficultto evaluate the performance of any mobility protocol or to devise any improvements.
(a) As far as I know, there has not been a systematic mobility system model that can analyze thesystem optimization techniques for cellular or IP-based mobility management protocols.
(b) Existing work does not provide a generalized mobility optimization framework that can sup-port horizontal and vertical handovers across administrative domains.
3
(c) There has not been any formal analysis of how mobility optimization might affect othersystem resources in the network.
This thesis proposal addresses the above issues. I analyze the basic operations associated with amobility event in detail. I develop a formal mobility systems model by representing the basic operationsassociated with a handoff as a series of discrete events. I formalize the associated states and transitionsin the form of a Discrete Event Dynamic Systems (DEDS) model. I then analyze the DEDS-orientedmobility model using Discrete Time Transition Petri nets. Based on the analysis of these propertiesassociated with a mobility event, I propose several systems optimization techniques for the basic oper-ations associated with the handoff event. I demonstrate these techniques by models, experiments, andnumerical analysis using a few network layer and application layer mobility protocols. I then applythese optimization techniques to a Timed Petri net model and compare with the experimental results. Ialso perform a tradeoff analysis of the utilization of system resources and handoff performance metricsobtained through these optimization techniques.
4 Systems Analysis of Mobility EventThis section analyzes the discrete operations of a mobility event that can help design a formal mobil-ity systems model. However, I limit the analysis of the mobility systems model to infrastructure-basedmobility only. Infrastructure-based mobility model assumes that the core layer 2 and layer 3 networkcomponents, such as access points, routers and servers are not mobile and only the last hop of the net-work keeps changing. Figure 1 shows the functional components of an infrastructure-based mobilityenvironment. It consists of layer 2 and layer 3 points of attachment, configuration agent, authenticationagent, authorization agent, and signaling agent that perform these primitive operations. During a mo-bility event, each of these network components participates in a distributed manner to take care of thehandoff related operations.
Backbone
Visited Domain
L2 PoAA
C
BD
CorrespondingHost
IPch
N2N1N1
N2
N1- Network 1 (802.11)N2- Network 2 ( CDMA/GPRS)
ConfigurationAgent
L3 PoAMobileHost
AuthenticationAgent
Authorization Agent
RegistrationAgent
RegistrationAgent
Home Domain
ConfigurationAgent
Authorization Agent
SignalingProxy
AuthenticationAgent
SignalingProxy
L3 PoA
L2 PoA
L2 PoAL2 PoA
L2 PoA
L2 PoA
L3 PoA
MobileHost
Figure 1: Functional components of a sample infrastructure-based mobility
In order to determine these abstract set of operations that constitute a handoff event, I investigatedand analyzed several cellular and IP-based mobility protocols at different layers. In Table 1, I summarizehow the various cellular and IP-based mobility protocols perform similar set of operations to support ahandoff event. Column headings of the matrix represent the fundamental operations that are part ofa handoff event for different types of mobility protocols represented in the row headings. I have alsoperformed a comparative analysis of application layer mobility and network layer mobility [29].
4
Table 1: Mapping of basic operations of a mobility event
MobilityType
AccessType
NetworkDiscovery
ResourceDiscovery
TriggeringTechnique
DetectionTechnique
Configuration Authentication Encryption BindingUpdate
MediaRerouting
GSM TDMA BCCH FCCH ChannelStrentgh
SCH TMSI SRES A3 DES MSCControl
Anchor
WCDMA CDMA PILOT SYNCchannel
ChannelStrength
Frequency TMSI SRES/A3 AES NetworkControl
Anchor
IS-95 CDMA PILOT SYNCchannel
ChannelStrength
RTC TMSI Diffie-Hellman/AKA
Kasumi MSCControl
Anchor MSC
CDMA1X-EVDO
EV-DO PILOT SYNCchannel
ChannelStrength
Frequency TMSI Diffie-HellmanCAVE
AES MSC PDSN MSC
802.11 CSMA/CA Beacon11R
11R802.21
SNR atMobile
ScanningChannelnumberSSID
SSIDChannelnumber
802.1XEAPoL
WEPWPA802.11i
Associate IAPP
Cell IP Any Gatewaybeacon
APbeacon
AP bea-con ID
GWBeacon
MACAddress
IKE AES RouteUpdate
Intermediaterouter
MIPv4 Any ICMPRA, FAadver-tisement
ICMP FA ad-vertisement,L2assisted
FA ad-vertisement
FA-CoACo-CoA
IKE PANAEAP
AES MIPRegistration
FA RFA HA
MIPv6 Any StatelessRAProactive
CARD802.21802.11R
RouterAdv.
RouterPrefix
CoA IKE PANAEAP
AES MIP Up-dateMIP RO
CH MAPHA
SIP-basedmobility
Any StatelessRAICMPRouter
802.21802.11R
L3RouterAdv.
RouterPrefixICMP
CoA AORRegister
IKEEAP InviteExchange
AESSRTP
Re-INVITE
B2BUA CHRTPTrans
4.1 Analysis of handoff componentsBased on the analysis of the basic operations, I categorized the handoff process primarily into six mainphases, namely, handover initiation, network discovery, network selection, network attachment, con-figuration, and media delivery. Each of these operations actually consists of several sub-operations.
1. Handover initiationDepending upon who initiates the handover and who has the primary control over the handover,a handoff process can be either mobile-initiated handover, network-initiated handover, mobilecontrolled handover or network-controlled handover [30]. During the first phase, the mobile orthe network node determines the need for the handoff based on the network condition, and initiatesthe trigger to start the handoff operation. For example, during a mobile-initiated handoff, based onSignal-to-Noise Ratio (SNR) or the quality of service of the ongoing application, the mobile makesthe decision about the impending handoff and starts the network discovery process to determinethe best available network to connect to.
2. Network discoveryThe second phase is the discovery phase, where the mobile or network discovers possible newpoints of attachment. This phase involves discovering both the neighboring networks and the re-sources within the network. Once the target network is discovered, several resource parameterswithin the target network are retrieved including channel number, bandwidth, encryption algo-rithm, authentication server, registration server, and configuration server. The resource discoveryprocess helps the mobile to configure with proper channel number and proper authentication pa-rameters in the new network so that it can communicate successfully. Jari et al. [31] providean overview of network discovery and selection, and its applicability to handoff. Based on thetype of access technology, the discovery process could be passive network discovery where the
5
node listens for network announcements or active network discovery where the node solicits thenetwork announcements. Based on the type of networks, the discovery process could involve dif-ferent operations, such as discovery of a new Location Area (LA) in case of GSM or discoveryof new Routing Area (RA) in case of GPRS. For IP-based networks, discovery process can spanacross all the layers and include cell, subnet or domain discovery. A domain is usually defined asan administrative domain.Each layer 2 access technology provides different means of discovering the networks and re-sources. As an example, GSM uses BCCH (Broadcast Control Channel), CDMA uses pilot chan-nel and 802.11 uses active scanning and passive scanning to discover the new point-of-attachment.Based on the type of access characteristics of the target networks, discovery of the appropriate net-work takes different amount of time. In case of IP-based mobility, some of the upper layer detec-tion mechanisms, such as foreign agent advertisement or router advertisement can help discoverthe new points of attachment at layer 3 and above.
3. Network selectionNetwork selection is a process by which a mobile node or a network entity analyzes the infor-mation discovered about its neighboring networks, and then selects a network to connect to. Theselection may be based on criteria such as required QoS, cost or user preferences. An appro-priate selection mechanism helps the overall resource optimization process that can increase theprobability of successful handoff to the target network.
4. Network attachmentAfter the mobile has selected the target network, it attempts to connect to the new network pointof attachment. Event notification, such as the availability of the new point of attachment or suddenlapse of an existing connection is usually provided to the upper layers so that any further handoffrelated functions can be expedited during a handoff event. The IETF is currently working on stan-dardizing a protocol for Detection of Network Attachment (DNA) [32] that involves mechanismsboth at layer 2 and layer 3 and can notify the upper layer about the network detection. The IEEE802.21 working group is currently defining several event service primitives. “Link Up” and “LinkDown” are examples of such primitives that can be used to provide the status of lower layer eventsto expedite the handoff process. For example, since the layer 2 association takes place before anyupper layer operations, it helps to send a layer 2 event notification, such as “Link up” to executethe upper layer mobility functions, such as layer 3 configuration. Signal-to-Noise Ratio (SNR)threshold can possibly generate such an event notification. Thus, an event notification from lowerlayer helps to establish a successful link in the new point of attachment in an expedited manner.
5. ConfigurationFifth phase is the configuration phase. This phase helps to prepare the re-connection path of themobile. During this phase, the mobile obtains a temporary identifier and establishes the mappingof its connection identifier with the appropriate network entity in the network. Configuration phasecan be categorized into the following sub-phases.
(a) Identifier configurationThis process allows a mobile to acquire a new temporary connection identifier either at layer2 or at layer 3 in the new point of attachment of the network. As an example, it could includeCare-of-Address (CoA) in case of IP-based mobility or TMSI (Temporary Mobile SubscriberIdentity) in case of GSM. Identifier configuration process involves obtaining a new identifier,testing the uniqueness of this identifier and finally assigning it to the mobile’s interface. Ingeneral, completion of these processes requires a series of signaling messages between themobile and a server in the network.
6
(b) RegistrationRegistration is a process of establishing the mapping between the permanent locator, suchas URI (Universal Resource Identifier) or home address and the temporary identifier, suchas care-of-address for proper location management function. An optimized or hierarchicalregistration process helps to expedite location management and provides faster delivery ofthe new data.
(c) Authentication, authorization and accountingThe authentication process allows the mobile to get access to the network resources. An au-thentication process involves a signaling exchange between the mobile and the authenticationserver in the network and generates the key that encrypts the data upon successful authentica-tion. Each mobility protocol uses different mechanism for authentication. As shown in table1, the mobile uses SRES and A3 algorithm for the authentication in GSM. For 802.11 accessnetworks, the mobile could use open system authentication, shared authentication WEP orstronger authentication 802.11i in layer 2. Fathi et al. [33] demonstrate how different au-thentication mechanisms affect the association and transport delay at layer 2. Similarly, layer2 access independent authentication protocols, such as PANA (Protocol for carrying Authen-tication to Network Access) [34] add delay during layer 3 authentication. Georgiades [35]shows that it takes up to 4 seconds to complete the authentication and authorization process.These operations could use a combination of EAP (Extensible Authentication Protocol) [36]over layer 2 for IEEE 802.1x-based authentication and EAP-TLS (Transport layer Security)[37] for layer 3 authentication. At the time of reconnection, re-authentication process adds tothe handoff delay. However, during inter-domain mobility, the authentication process is typ-ically followed by authorization process that involves interaction with authorization server,such as AAA.
(d) Security associationA security association can be defined as a secure channel between two endpoints that appliesa security policy and keys to protect information. Before a new communication path is es-tablished between the end-points, mobile node needs to authenticate itself and then establishsecurity association with other network nodes. Establishing a security association always in-volves a key distribution procedure that includes exchange of messages between the mobileand any centralized server. Upon a successful derivation of the key, the mobile can encryptthe data for protection. Security association may take place at several layers of the protocolstack. For example, in 802.11 environment, a 4-way handshake is needed between the mo-bile node and the access point to generate the PTK (Pairwise Transient Key) that can be usedfor encryption. In an IP-based network, ISAKMP (Internet Security Association and KeyManagement Protocol) [38] defines the mechanism for establishing the security association.
(e) Binding updateBinding update is the process by which a mobile can update its new network identifier sothat the data can be rerouted to the new destination. As the mobile connects to a new pointof attachment and obtains a new temporary network identifier (e.g., TMSI in GSM, COAin MIPv6, FA-COA in MIPv4) in the new network, it needs to update the correspondenthost or home agent so that the packets can be routed to the new destination. This processassociates the new network identifier with the permanent identifier of the mobile. Until thereassociation of the new identifier is complete, the transient data still goes to the old networkand is considered lost in the absence of any optimization mechanism, such as bufferingor packet forwarding. In some cases, this binding update needs to be authenticated. Forexample, MIPv6 allows return routability procedure and adds two additional messages, suchas CTI (Care-of-test Init) and HTI (Home-test Init) to obtain the binding key so that bindingupdate can be authenticated as well. This process contributes to the additional delay for thebinding update procedure.
7
6. Media reroutingA media rerouting phase is considered to be the last phase during the handover process. Thisphase involves re-routing of the media so that the delivery of data changes from the old path to thenew path according to the pre-defined service guarantee.Once the binding update is complete, data from the correspondent node gets routed to the mobile’snew location. Media delivery can take place in several ways. In one way, the media delivery cantake place using the direct path between the CN and the MN. In second scenario, the media isdelivered using an indirect path and uses network entity such as home agent. In both cases thein-flight data can be captured and redirected to the new point of attachment. Media rerouting pro-cess may include several elementary operations, such as encapsulation, decapsulation, tunneling,buffering, and store-and-forward technique. During the media re-routing process, transient datamay get lost or may get delayed because of these operations. There is certain overhead associatedwith encapsulation, decapsulation and tunneling operations. Thus, there is a need to optimizethese operations to ensure that the media delivery is not delayed. Optimization techniques for themedia delivery are often defined as route optimization methodologies. I describe some of thesetechniques in Section 6. As an example, operations in the network, such as buffering or forward-ing help to reduce packet loss but add delay to the packet traversal. Thus, the buffering periodneeds to be adjusted to compromise between packet loss and one-way packet delay.
4.2 Effect of handoff across layersIn this section, I illustrate how these basic handoff operations span across multiple layers in an IP-basedenvironment. Based on the type of mobility, a subset of operations are executed at each layer and theoverall handoff delay is affected by delays at all layers. For example, during an intra-domain movement,the mobile is not subjected to the delay due to authorization unlike inter-domain mobility. Following isa description of how optimization techniques can be applied to each of these basic operations at eachlayer.
Layer 2 delay: In a typical 802.11 environment, channel scanning, probing, authentication, and as-sociation are the basic functions that contribute to the delay before a mobile completes the networkattachment at layer 2. Scanning is considered to be a discovery process in layer 2. Encryption and userauthentication using WPA (Wi-Fi Protected Access) in conjunction with 802.1X and EAP (ExtensibleAuthentication Protocol) contribute to the additional delay because of associated 4-way handshake be-tween the mobile and the access point. Mishra et al. [39] provide a comprehensive delay analysis of thebasic operations associated with a layer 2 handoff.
I have taken measurements to study the layer 2 handoff related operations for two different oper-ating systems, Linux and Windows, and used different layer 2 drivers, (Aironet, Orinoco, DLink, andCentrino). A thorough analysis of the layer 2 handoff event shows that scanning and probing operationscontribute to most of the delay. For example, in our current experimental setup, using active scanningwith Orinoco driver in a Linux environment, it takes almost 100 ms for probing action to complete. Thisis followed by layer 2 open authentication and association that take about 2 ms and 20 ms, respectively. Ihave also experimented with layer 2 authentication, such as IEEE 802.11i, and have found the EAP delayand 4-way handshake delay to be 79 ms and 616 ms for non-roaming and roaming cases, respectively.
Layer 3 delay: In an IP-based environment, a layer 3 association goes through several basic opera-tions, such as discovery of layer 3 point of attachment (e.g., subnet), new IP address acquisition, du-plicate address detection, neighbor reachability, local access authentication, and authorization. Each ofthese operations involves a number of message exchanges between the mobile and other network entity.
8
By means of periodic router advertisement, the mobile can discover the new layer 3 point of attachment.There are several protocols, such as DHCP [40], DHCPv6 [41], PPP [42] or stateless auto-configuration[43] that are used to take care of the IP address acquisition process. Table 2 shows IP address acquisitiondelays as measured by different configuration protocols. Time taken by each of these mechanisms differsdue to various factors, such as variation in the number of signaling messages to complete the addressacquisition, duplicate address detection technique. However, in a typical inter-domain mobility scenario,there are additional operations, such as re-authentication, re-authorization and profile verification by theAAA servers during the handoff. In some cases, the AAA-related messages traverse all the way to thehome AAA server before the network service is granted to the mobile in the new network. Thus, it isdesirable to have an optimized method of interaction between the mobile and AAA server during thehandoff. After a layer 3 identifier is reconfigured, other layer 3 functions, such as the binding updatefrom the mobile and media redirection at the correspondent host contribute to the additional delay.
Table 2: IP address acquisition delay
Layer 3 configu-ration methods
DHCP w/ARP
DHCPw/o ARP)
IPv6Sateless
DHCPv6Stateful
PPP FACoA
ZeroConf
Static Proactive
Address acquisi-tion delay
4 s 400 ms 160ms
500 ms 8 s 2 s 5 s 100ms
4 ms
Application layer delay: Application layer delay is mostly attributed due to the operations such asbinding update delay at the application layer, processing delay at the end hosts, registration, upper layerencryption, such as TLS (Transport Layer Security), and SRTP (Secured RTP). SIP-based mobility isa typical example where binding update is done at the application layer. This contributes to the delaybecause of three-way-exchange between the mobile and the correspondent host during rebinding process.
Table 3 shows the summary of how the basic operations of a mobility event described in Section 4.1are taken care of across different layers in an IP-based environment.
Table 3: Mapping basic handoff operations across layers
Layers Basic handoff operationsDiscovery Authentication Security
AssociationIdentifierConfiguration
AddressUniqueness
BindingUpdate
MediaRouting
Layer 2 Scanning Open authEAPOL
4-wayhandshake
ESSIDBEACON
MACAddress
Layer 2UpdateCache
IAPP
Layer 3 Routeradvertisement
L3EAPIKEPANA
IPSEC DHCPStateless
ARPDAD
UpdateHACN
EncapsulationTunnelingForwarding
ApplicationLayer
AAAdomaindiscovery
S/MIME TLSSRTP
URI-IPMapping
Registration SIP Re-INVITE
DirectRouting
5 Modeling Mobility as Discrete Event Dynamic SystemIn order to understand the interaction between basic operations of handoff and investigate the perfor-mance metrics, such as handoff delays and resource utilization during the handoff operations, it is im-portant to design a formal system model that can analyze the important system properties of the mobilityevent. In this section, I develop the mobility system model by including the state transition associatedwith the basic operations during the handoff. This model can help evaluate the efficiency of any mobil-ity protocol and analyze the tradeoff between performance metrics and resources when a mobility eventdemonstrates parallelism, optimistic or speculative operations.
A few related efforts have attempted to model certain aspects of mobility management. Marshan etal. [44] have used a Petri net-based model to analyze the performance characteristics of wireless Internet9
access for GSM and GPRS systems. Amadio et al. [45] have modeled IP mobility using a processcalculus approach and have applied this to a specific protocol, such as Mobile IPv6. The process-calculi-based method has actually used a software agent approach for modeling the mobility event. Tutsand Sokol [46] provide a Petri net-based performance evaluation of bandwidth partitioning meant forwireless call level. Molina-Ramirez et al. [47] and Jaimes-Romero et al. [48] have used a Petri net-based approach to model resource management in cellular systems. However, none of these papers haveactually attempted to model the systems aspects of a mobility event. None of these models were intendedto systematically analyze the elementary operations involved with a mobility event or serve as a basisfor optimizing the mobility event. Based on the analysis of the mobility event, I develop a formal modelthat can be applied to the analysis of any type of mobility protocol.
The mobility event can be viewed as the perturbation to the steady state of a communicating nodethat may affect different layers in the protocol stack. As a communicating node is subjected to handoff,it goes through a series of sequential discrete states before it attains a steady state by returning to theconnected state. Each of the basic operations described in Section 4 can be modeled as a transition eventwhere the mobile moves from one state to another. As a result of a series of state transitions involvedduring the mobility event, a mobile’s communication is interrupted because of the delay associated witheach transition. Discrete Event Dynamic System (DEDS) corresponds to a type of system where the statespace is discrete and state changes are driven by external or internal events. A mobility event can thusbe modeled as a discrete event dynamic system, since it exhibits concurrent, sequential, and competitiveactivities (i.e. two operations competing for the same resources) among many of the several handoffoperations that are part of the mobility event.
I considered several tools as possible candidates to model and analyze the DEDS-based mobilityevent. These include software language, such as Esterel [49], Markov chains, queuing theory, mini-max algebra [50], GNATT charts [51], and Petri nets. I finally chose Petri net as the tool to model themobility event for the following reasons. Petri nets can be used to model properties such as processsynchronization, asynchronous events, sequential operations, concurrent operations, and conflicts or re-source sharing. Timed Petri nets provide a uniform environment for modeling, design and performanceanalysis of discrete event systems. The graphical representation makes Petri nets intuitively very ap-pealing to represent any temporal events. As a mathematical tool, a Petri net model can be described bya set of linear algebraic equations or other mathematical models reflecting the behavior of the system.Several software tools are available (e.g., TimeNet [52], SPNP [53], UltraSAN [54], TOMSPIN [55]) toautomate the analysis of the Petri net model and obtain performance evaluation. Timed Petri net modelcan also be used with other existing technique, such as simulated annealing to conduct a tradeoff analysisbetween different performance metrics and systems resources.
Mobility event modeling can be viewed as analogous to modeling of the Flexible ManufacturingSystem (FMS) as both exhibit a series of sequential operations. Zuberek et al. [56] model and analyzethe simple schedules for manufacturing cells. Similar techniques can be applied to conduct performanceanalysis for the mobility systems model. I use Deterministic Time Petri net [57] to model the mobilityevent and derive the relevant optimized models by applying the appropriate concurrent and proactivemechanisms. I represent the basic operations in terms of places and transitions, and then validate theoptimization techniques associated with each of these operations.
Figure 2 shows a high level view of how discrete states associated with a mobility event can be rep-resented in a formal framework by using a Timed Petri net approach [57]. In this figure, each place (P i)represents various stages of the mobility event and the transition (ti) represents the time taken to com-plete different set of operations between the stages. Each of these stages can be considered as multiplesub-systems of the system representing the mobility event. Each of these transition processes can actu-ally consist of several sub-processes. For example, as shown in Figure 2, transition process t1 (NetworkResource Discovery) consists of several sub-processes that are connected by several sub-transitions, suchas t01, t02, t03, and t04. Table 4 provides a description of places and transitions associated with Figure2. Each transition process has been representative of a multiple of chosen time unit.
The Petri net model representing the general mobility systems is actually a decision free Petri net,10
ConnectedP8
Disconnected
NetworkDiscovered Network
Selected
MobileConfigured
AuthenticatedSecurityAssociationEstablished
Updated
NetworkResourceDiscovery Network
SelectionDetectionProcess
NetworkConfiguration
AuthenticationProcess
Binding Update
MediaForwarding
P1 P2 P3
P4P5P6
P0
t1 t2 t3
t4
t5t6t7
P7
Intra-domainBinding update
t8t9
BufferingRedirection
t0
p01
p02
p03
t01
t02t03
t013
t04
Security AssociationProcess
Figure 2: A generalized Petri net model for a mobility event
Layer 2
Layer 3
ApplicationLayer
Layer 2Event
Layer 3Event
Layer 4Event
Enter Mobility Event
Leave layer 2
Disconnected
Leave layer 3
Leave mobilityevent
Enter layer 3
Enter layer 2
Layertransition
Connected
SNR goes belowa threshold
State
Service
discovery
Scanning is performed
NetworkSelected
L2 authenticationperformed
Authenticated
Location
L3 discovery
L3 addressacquisition
DAD
configuration
L3 authenticationperformed
authenticated
Extraneous action
buffering
BU performed
Forwarding
Media redirection
Figure 3: Layered modeling with Petri net
where a minimum cycle time is an indicator of maximum performance. The cycle time is represented asC = max Tk/Nk:k=1,2,3...q, where Tk = sum of the execution times of the transmissions in circuit k and(Nk) is the total number of tokens in the places in circuit k and q is the number of circuits in the net. Incase of a system model representing a mobility event, these values can vary depending upon the numberof transitions and sequence of transitions involved during a cycle.
Figure 3 shows a subnet level view of a mobility systems model as represented by a modular Petri net.It includes the state transitions at each layer and illustrates the component-level interaction in each layerand across layers. Several subprocesses in each of the layers, such as discovery, layer 2 authentication,layer 3 address acquisition, layer 3 authentication, security association, and binding update are illustratedhere. The reduction of the number of transitions, minimization of time during each of the state transition,parallelization of many of the state transitions within each layer and across layers, and reduction ofnumber of states will contribute to the overall optimization of the handoff process.
Figure 4 shows the modeling of the mobility event using TimeNet software. It takes into accountthe hierarchical feature of Petri net [58] and shows the granular level interaction between different stateswithin a mobility event. A Petri net-based model can thus be used to analyze various types of mobility
11
Table 4: Description of Places and TransitionsPlace DescriptionP0 Mobile is in disconnected stateP1 Network and Resources DiscoveredP2 Target Network SelectedP3 Mobile is configured and registeredP4 Mobile is authenticatedP5 Security Association is establishedP6 Binding Update is completeP7 Intra-domain binding update completeP8 New media reaches the mobileTransition Description Time De-
layt0 Mobile gets disconnect trigger 1tt1 Mobile discovers the network and resources at new PoA 2tt2 Mobile selects the network 3tt3 Mobile goes through configuration and registration process 4tt4 Mobile goes through authentication process 3tt5 Mobile goes through key derivation and security association process 2tt6 Mobile goes through binding update process 6tt7 Mobile goes through hierarchical binding update 5tt8 Media gets redirected to the mobile 2t
event, such as intra-subnet, intra-technology, inter-subnet, and inter-technology handover.In Petri net models, a cycle time represents the associated delay during the handover process that
can be related to overall overall efficiency of the system. Ordering of execution of the processes thatconstitute a mobility event plays an important role in determining the cycle time. Figure 5 illustratesPetri net representation of possible sequence of execution for a pair of processes, such as Pa and Pb thatare part of a mobility event. Thus, ordering of sequence of execution of sub-tasks during a mobilityevent affects the overall cycle time although amount of system resources expensed may vary.
Thus, several optimization techniques can then be applied to the generalized mobility model andcorresponding optimized models can further be derived. These techniques can be applied to the processeswithin the overall system or in a hierarchical manner between each of these sub-processes. Performanceevaluation can then be performed for each of these optimized models and be compared with the non-optimized models. Most of the optimization techniques that I have developed can primarily be mappedinto sequential, concurrent, and proactive modes of operation in a Petri net model.
The systems performance of a Petri net model representing a mobility event can be verified in severalways. In one scenario, minimum cycle time can be obtained from the Petri net model using the numberof circuits, number of transitions, and timing associated with each transition. Table 5 shows how theoverall cycle time is affected when some of the optimization techniques, such as hierarchical bindingupdate, proactive discovery and configuration, and anchor-based security association are applied to ageneralized Petri net system mobility model shown in Figure 2.
Similarly, in order to verify the systems performance of a mobility event demonstrating a specificoptimization methodology and to determine if the system satisfies the desired threshold for cycle timeone can generate a Place matrix (P), Transition matrix (Q). Entry (A,B) in the matrix P equals “x” ifthere are “x” tokens in place A, and place A is connected directly to place B by a transition. Entry (A,B)in the transition matrix “Q” equals ti if A is an input place of transition ti, and B is one of its outputplaces. Entry (A,B) contains symbol “w” if A and B are not connected directly. Given a threshold valueof cycle time C, one can generate a distance matrix CP-Q. Then using the Floyd [59] algorithm, one canthen determine matrix S. By inspecting the values of the diagonal elements of matrix S, it is possible
12
NETWORKSELECTED
CLIENTCONFIGURED
AUTHENTICATED
ENCRYPTED
HAUpdated
MNConnected
Tkey
TencryptTHAforward
MNDisconnected
Tdiscon
T01 T02 T03 T04
Chanscanned L2OpenAuthL2Assoc EAP
T2
WPA
T3T4
T5
L3routeradvT6
t1
NETWORKDISCOVERED
T1
AddrAcquisition
T7
DuplicateAddress
T8
KeymgtT9 SAdone
T10
COTIHOTI
T11
T12
BUtoHA T13T14
BUtoCN T15CNupdated
TCNforward
MNreceivesMedia
BUtoAnchorHA T17
T16
T18ANhorHAupdateTTransientTrf
T19
Figure 4: Petri net representation of mobility event using Time Net
to determine if the system satisfies the desired system performance. A set of matrices that represent asample Petri net model consisting of 4 places with a given value of C = 3 are shown in Equation 1. In thisspecific example, all the diagonal entries of matrix S are non-negative. Thus, the system performancejust meets the threshold requirement of C = 3. By way of this analysis, one can easily determine if aspecific Petri net model can meet the desired system performance.
Table 5: Cycle time for Petri net with mobility optimizationsType ofOptimization
Loops in the state transition path Di Ni Di/Ni
Cycle
TimeNo optimization p0t1p1t2p2t3p3t4p4t5p5t6p6t7p10 24t 1 24t
Hierarchicalbinding update
p0t1p1t2p2t3p3t4p4t5p5t8p7t9p10 19t 1 19t
Proactive p0t9p10 2t 1 2t
Maintain securitybinding
p0t1p1t2p2t3p5t6p6t7p10 19t 1 19t
P =
0 1 0 0
0 0 1 0
0 0 0 1
1 0 0 0
,Q =
w 2 w w
w w 3 w
w w w 4
3 w w w
,CP−Q =
∞ 1 ∞ ∞
∞ ∞ 0 ∞
∞ ∞ ∞ 1
0 ∞ ∞ ∞
,S =
0 1 1 1
0 0 0 1
0 1 0 1
0 1 1 0
(1)
6 Proposed Mobility Optimization TechniquesIn this section, I describe some of the key optimization techniques that I have developed to optimizeseveral basic operations of a mobility event and describe the associated general principles.13
Pa Pb
Pa Pb
PaPb
Pa
Pb
tc
Pa
Pb
pbpa
tc
tc
ta tc tb
ta tbta
tc tb
tc ta
tb
ta
tb
pa starts before pb
pa meets pb
pa overlaps pb
pa during pb
pa starts pb
pa finishes pb
pa starts with pbpapb
tctb
tc tata
tb
Figure 5: Petri net mapping for sequence of events
These techniques are based on some very fundamental rules of optimization, such as reduction ofsignaling messages during the basic operation, minimizing the traversal distance of the data, reduction ofdata and signaling overhead, minimization of look-up cost, caching, parallelization of sequential opera-tions, proactive operations, cross layer triggers, and localizing the mobility related operations. I applythese optimization techniques to Petri net-based mobility model and perform experimental evaluationon each of these techniques. Then I experiment with a few of these proactive techniques to perform anoverall systems evaluation.
In order to experiment with the key optimization techniques, I have implemented an Internet mobilemultimedia testbed where I can demonstrate all the functional components of the mobility event. I haveimplemented configuration agent, signaling agent, mobility agent, home agent, authentication agent,and authorization agents using the IETF-based protocols, such as DHCP, SIP, MIP, PANA, and Diameterover heterogeneous access networks, such as 802.11 and CDMA. In particular, I describe the detailsof the implementation of the multimedia testbed in [60]. I describe below the optimization techniquesassociated with some of the primitive operations that are part of a handoff event.
6.1 Application layer discoveryIn this section, I describe an application layer network discovery mechanism that can discover the net-work elements in the neighboring networks independent of the underlying access technology. Using thisdiscovery technique, mobile can proactively discover the network resources in the target network andoptimize the handoff process.
As discussed in Section 4, experimental results show that network discovery and resource discoveryprocesses in IEEE 802.11 networks contribute to a large amount of delay. There are related work thatattempt to reduce the network discovery time in 802.11 environment. Shin et al. [61] adopt a selectivescanning and caching strategy to reduce the 802.11 handover latency. However, this method is moreapplicable to an environment where the mobile has associated with the neighboring APs in the past,and cannot be applied if the target access point is a new AP. Montavont et al. [62] propose a periodicscanning method, where the mobile does the scanning in different channel periodically and builds upthe list of APs. However, this mechanism seems to generate more traffic, and as a result consumes morepower. Velayos et al. [63] provide techniques to reduce the layer 2 discovery process by reducing thebeacon interval time and performing the search phase in parallel with data transmission. Brik et al. [64]propose to use a second interface to scan while communicating with the first interface, thereby avoidingscanning delay during communication. There are few task groups within IEEE 802.11 standards that
14
propose new network discovery mechanisms at layer 2 and application layer. The IEEE 802.11u [20]working group proposes methods of network selection with other external networks, such as cellular.The IEEE 802.11k [19] working group proposes three different ways of discovering networks around anaccess point. However these discovery mechanisms are limited to 802.11 access only and do not work fornon-802.11 access. Most recently, Andrea and Schulzrinne [65] have developed discovery mechanismsusing cooperative roaming techniques that are suitable to work in an infrastructure less environment.
As part of my initial work, I have developed an access independent information server-based discov-ery mechanism to help discover the network parameters and resources of the target network [66]. Unlikethe existing network discovery mechanism, the application layer discovery mechanism does not dependon discovery mechanism associated with any specific access technology. This technique can be appliedto both infrastructure-assisted and end-system assisted scenarios.
I have described an analysis of how this discovery mechanism can be effective in a collaborativeenvironment using end-system assisted approach [67]. As part of infrastructure-assisted scheme, theinformation server stores the details of the networks and the associated resources in a generic formatin an access independent manner that can be queried by the mobile client at any time. The client com-municates with the information server and discovers the neighboring network elements, such as accessrouter, authentication agent, configuration agent and authorization agent, and communicates with theseentities prior to its handover to these networks. By discovering prior to handoff, the mobile keepsthe MAC address and channel number in its cache and thus avoids the complete scanning procedureduring 802.11-based handover. Prior discovery of routers and authentication servers also helps to com-plete other handoff related functions, such as authentication and configuration prior to handoff [68].The proposed information server-assisted discovery technique has been adopted as one of the discoverymechanisms for the Media Independent Information Server (MIIS) function of IEEE 802.21. Proactivediscovery of this information expedites the handoff operation. I have implemented the information ser-vice using RDF (Resource Description Framework). Evaluation of a complete system using networkassisted discovery scheme is described in Section 7.11.
6.2 Network layer assisted pre-authenticationIn general, authentication and authorization take place in the target network after the mobile moves tothe new network. For example, in IEEE 802.11-based networks, the authentication mechanism requiresIEEE 802.1X message exchange with the authenticator in the target network, such as an access pointthat can initiate an EAP exchange with the authentication server. Following a successful authentica-tion, a four-way handshake with the wireless station derives a new set of the session keys to encryptthe data. The handover latency introduced by this authentication mechanism has proven to be largerthan what is acceptable for some handover scenarios involving inter-domain handover. Hence, improve-ment in the handover latency performance due to authentication procedures is a necessary objective forsuch scenarios. Various standard organizations are developing access specific techniques to reduce theauthentication delay. But these efforts are limited to specific access technologies.
The IEEE 802.11f, a trial use document has defined context transfer and chaching mechanism totransfer some 802.11i keying related information between the neighboring APs. However, it has been ad-ministratively withdrawn since 2006. The IEEE 802.11i defines a pre-authentication mechanism that canbe used in 802.11 variant wireless networks. This mechanism allows mobile devices to pre-authenticateusing EAP to one or more target authenticators over the wired medium, by way of the current authen-ticator. The IEEE 802.11r has been working to define “Fast BSS” transition mechanisms involving adefinition of key management hierarchy and setup of session keys before the re-association to the tar-get AP. These mechanisms are however defined for IEEE 802.11 technologies and are only applicablewithin a specific mobility domain and fall short when it comes to inter-access technology handovers.They also require L2 (e.g., Ethernet) connectivity for transfering encapsulated signaling to the target AP.The problem of applying link-layer handoff optimization mechanisms between different subnets has alsobeen addressed by Barg et al. [69] and Georgides et al. [35]. However, most of the solutions are basedon context transfer mechanisms where the optimization is achieved by transferring the security context
15
Table 6: Network layer assisted layer 2 preauthentication
TypesOfAuthentication
IEEE 802.11iEAP/TLS
Post-authentication
IEEE 802.11iPre-authenticaton
Network LayerAssisted Layer 2
Pre-authenticationOperation Non
RoamingRoaming Non
RoamingRoaming Non
RoamingRoaming
Tauthentication 61 ms 599 ms 98 ms 638 ms 177 ms 831 msTconfiguration
2 AP– – – – 16 ms 17 ms
Tassociation+4 way handshake
18 ms 17 ms 16 ms 17 ms 15 ms 17 ms
Total 79 ms 616 ms 114 ms 655 ms 208 ms 865 msTime affectinghandover
79 ms 616 ms 16 ms 17 ms 15 ms 17 ms
(keys and related parameters) created by the client and previous AP to new AP in the target subnet. Also,from the security perspective, it is not always a good idea to transfer the cryptographic keys betweendifferent network entities beloging to two different domains. Housely and Bernard have also raised awarning on security context transfer [70].
I have proposed a network-layer assisted mechanism [71], [72] that can take care of many of thedrawbacks of the existing approaches and assist with layer 2 pre-authentication independent of the ac-cess technology. This mechanism is based on the following basic principles: reduction of authenticationrelated signaling delay prior to mobile’s handoff to the new network and reduction of generation ofkeys that help to encrypt the data. In a typical inter-domain mobility scenario, an authentication pro-cess is followed by an authorization process. In addition to reducing the delay due to layer 3 relatedauthentication and authorization, these proposed mechanisms can reduce authentication delay at link-layer when existing link-layer handoff optimization mechanisms cannot be applied for handoff casesinvolving inter-domain and inter-access technologies. A successful authentication prior to handoff leadsto proactive configuration and establishment of security association between the mobile and networkelements in the target network.
During the proactive authentication phase, the mobile can initiate the link-layer security associationin the target network. After the mobile node chooses a specific candidate network as the target network,and switches to a new point of attachment in the target network, it just needs to execute 4-way handshakeusing the previously generated PMK (Pair-wise Master Key) and establishes PTKs (Pair-wise TransientKeys) and GTKs (Group Transient Keys) that protect link-layer packets between the mobile node andlayer 2 point of attachment. No additional execution of EAP authentication is needed here.
I have experimented with the proposed proactive authentication techniques to optimize both layer3 and layer 2 delays in IEEE 802.11 environment using PANA [34] and Diameter [73]. I have demon-strated the effect of pre-authentication using both application layer and network layer mobility protocols.Details of layer 2 authentication bootstrapping and experimental results are explained in [71]. Table 6provides the results from network layer assisted layer 2 pre-authentication. These results illustrate hownetwork layer assisted layer 2 preauthentication can provide comparable results with 802.11i-based pre-authentication but can support inter-domain mobility. Detailed results showing how pre-authenticationcan inter-work with other required handoff related operations for a composite system is provided inSection 6.11.
6.3 Anchor assisted security associationDuring the mobile’s repeated handoff security association between the mobile and communicating hostneeds to be re-established because the end-point identifier keeps changing. This rebinding process re-quires exchange of messages to derive the new key and processing at the end hosts and thus contributes tothe added delay. Miu et al. [74] describe an architecture that helps to maintain security association when
16
the mobile moves between public and private networks. However, this solution is limited to movementbetween similar kinds of networks (e.g., 802.11b). Rodriguez et al. [75] introduce the concept of mobilerouter where the end clients with multiple access technologies connect to the mobile router’s downlinkinterface. In this case, the end clients do not change their IP addresses, rather the mobile router keeps onchanging the external IP addresses as it moves around and connects to different access networks, such asGPRS, CDMA and 802.11b. The router uses NAT (Network Address Translator) functionality to shieldthe clients from re-initiating the sessions.
I have designed an anchor assisted mechanism that maintains the security association during thehandoff. The key principle introduced by this technique is to maintain the security association with theend client even when the end-point identifier changes. This avoids the delay due to re-establishmentof the security association during the mobile’s handoff. I have experimented with this technique usingboth network layer mobility and application layer mobility protocols. In the experiment, a mobile withtwo interfaces moves back and forth between an enterprise network equipped with 802.11, a cellularnetwork with CDMA1XRTT access, and a hotspot equipped with 802.11. By introducing an anchor suchas a secondary home agent in the network, one can achieve secured seamless communication withoutthe need to re-establish the IPSec [76] tunnels during each subnet move. I have experimented withthis technique with both CBR (Constant Bit Rate) traffic (audio) and VBR (Variable Bit Rate) traffic(video) and have analyzed the packet loss, delay, and jitter during the handoff. In the absence of suchan optimized scheme, the mobile experiences packet loss due to the delay associated with IPSec tunnelsetup and tear-down every time it changes its point of attachment. Without any optimization, layer2 configuration took about 10 seconds in a CDMA network, layer 3 address acquisition took about 3seconds in 802.11 network. Binding updates, such as both external and internal MIP registrations, tookabout 300 ms and 400 ms to complete. IPSec-based security association took about 6 seconds. As themobile moved back to the home network, it took around 200 ms for mobile IP de-registration. Theseresult in degradation of real-time services due to associated delay and packet loss. However, using acombination of anchor-based security association technique that helps to maintain the security bindingand a make-before-break technique, I could obtain zero packet loss during the handover from 802.11networks to cellular and vice-versa. Although there was no packet loss in the optimized case, the mobilereceived a few out of order packets during its movement back from cellular network to 802.11 networkas the transit packets in slow cellular link arrived later than the initial packets that arrived via 802.11interface.
This optimization technique takes advantage of two fundamental principles of optimization, such asreducing the number of signaling exchange and hiding the change of end-point address. I describe thedetails of the implementation and experimental verification in [77].
6.4 Layer 3 ConfigurationBefore assigning the IP address the client usually performs duplicate address detection (DAD) by way ofARP (Address Resolution Protocol) or Neighbor Discovery in IPv4 and IPv6 networks, respectively. Forexample, for an IPv4-based network, this detection procedure may take up to 4 seconds to 15 seconds[78]. DAD-related delay for stateless address configuration of IPv6 address identifier may take up to1500 ms and depends on the random value that determines Neighbor Solicitation interval.
For IPv6 networks, Moore et al. [79], Han et al. [80] propose some of the techniques needed to carryout DAD (Duplicate Address Detection) optimization for the IPv6 clients. Optimistic DAD [79] ensuresthat the probability of address collision is not increased and thus improves the resolution mechanismsfor address collisions. There are few proposals in the IETF, such as Passive DAD [81] and DHCP rapidcommit option [82] that try to expedite the IP address acquisition for IPv4 networks.
Initially, I investigated the effect of layer 3 configuration on the handoff for both IPv4 and IPv6networks. Then I proposed both reactive and proactive techniques that can expedite part of layer 3 con-figuration process. As part of my investigation into layer 3 optimization techniques using DHCPv4 andMIP, I have verified that with ARP check enabled, the IP address acquisition took an average of 15 sec-onds, but when the ARP check is suppressed average time taken for IP address acquisition was 436 ms. I
17
conducted several experiments to analyze the effect of two factors, such as Duplicate Address Detection(DAD) [43] and Neighbor Unreachability Detection (NUD) [83] on the disruption of real-time voicetraffic over IPv6 network using both Mobile IPv6 from USAGI [84] and SIP-based terminal mobility. Ienabled aggressive router selection procedure in the kernel module that helps the mobile to bind to thenew router quickly enough without doing a Neighbor Unreachability Detection [83]. The experimentalresults shown in Table 7 demonstrate how DAD- and NUD-related delays affect both signaling and me-dia redirection delay. However, by avoiding DAD and adopting aggressive router selection technique toreduce the effect of NUD, I could reduce the signaling delays to 200 ms and media delays to less than500 ms [85] for SIP-based mobility.
Table 7: Effect of duplicate address detection (IPv6) on handoff
Handoff Case Signaling Delay (ms) Media Delay (ms)SIP w/DAD
SIP w/oDAD
MIPv6 w/DAD
SIP w/DAD
SIP w/oDAD
MIPv6 w/DAD
H12(Home-Visited 1)
3829 171 2 3854 421 21
H23(Visited 1 – Visited2)
3932 161 2 4188 419 30
H31(Visited 2 – Home)
1935 161 1 1949 408 25
6.4.1 Router assisted duplicate address detection
I have designed and implemented a router assisted duplicate IP address detection mechanism that reducesthe layer 3 configuration time. The router typically keeps the list of IP addresses configured in a specificsubnet in its neighbor-cache. As part of this scheme, the router provides the list of the IP addresses inuse in the neighboring subnets periodically in a scoped multicast address.
Thus, a mobile can obtain this information from the router without performing an ARP. I explainthe details of the proposed duplicate address detection mechanism in [86]. Unlike other approaches,the proposed approach does not need any new element in the network and does not need changes in theDHCP server. However, there is a tradeoff between the frequency of router advertisement and the loadon the network. This technique also needs some modification on the router.
6.4.2 Proactive address caching
I have designed a proactive configuration technique that can work independently or in conjunction withpre-authentication mechanism. It uses the general principle of proactive caching technique. A proac-tive configuration mechanism consists of several steps, such as proactive address acquisition, proactiveduplicate address detection and proactive address resolution.
Although FMIPv6 [3] provides a method of proactive IP address acquisition, by obtaining the routerprefix from the next access router, it expects that the adjacent routers need to cooperate and discovereach other. Thus, this mechanism may not work for inter-domain mobility.
My proposed technique is client assisted, and can be applied to both intra-domain and inter-domainmobility scenario. In the proposed technique, the client obtains the IP address of the target networkwhile the mobile is still in the current network. It assigns this proactively obtained address to a virtualinterface and performs a subsequent proactive binding update. Alternatively, it can store it in the localcache and assign the address later on. This avoids the delay due to signaling messages caused during theaddress acquisition process.
During IP address acquisition process, duplicate address detection is performed ahead of time whilethe mobile is in the previous network. This is usually performed by DHCP relay that can colocate withthe next target router. In case of stateless address configuration, DAD is performed over the proactive
18
tunnel and with the help from the target router. During the process of pre-configuration the mobile canalso install the ARP cache of the neighboring first hop router with whom it is likely to communicate afterattaching to the new network. Having prior knowledge of IP address-to-MAC address mapping both theneighboring first hop router and the mobile do not need to discover each other at layer 2 (i.e., addressresolution) after the mobile moves to the new network.
I have demonstrated proactive address acquisition for both IPv4 and IPv6 networks using PANA.Independent of pre-authentication mechanism, I have also used stand-alone protocols, such as GIST(General Internet Signaling Transport) [87] and IKEv2 [6] to configure the mobile proactively. I havedescribed the details of this technique in [88].
It is inferred from these experiments that the number of messages passing between the client and theserver during IP address acquisition, processing time at the end systems, and network load are some ofthe key factors that contribute to the layer 3 configuration delay. Proactive caching of the IP address atthe client and router or server assisted proactive duplicate address detection technique reduce the layer3 address acquisition delay.
6.5 Route optimizationTriangular routing, encapsulation and decapsulation processes associated with any mobility protocolcontribute to the performance degradation because of the associated transport delay for signaling, data,and encapsulation overhead. Route optimization is the process of optimizing the route between the com-municating hosts by eliminating encapsulation and decapsulation processes and maintaining the directroute. The IETF has addressed these issues by proposing route optimization support for MIPv6 [3].However, route optimization for MIPv4 never got standardized but various forms of route optimizationhave been proposed by others [89]. In this section, I describe few route optimization mechanisms thatI have developed and illustrate the experimental results. These mechanisms optimize the signaling anddata route by maintaining the direct path between the communicating hosts and avoid any associatedencapsulation overhead. Since MIP inherently suffers from this route optimization problem, I experi-mented with a few of these optimization techniques and then compare the results with that of MIPv4.
6.5.1 Maintain direct path by application layer mobility
SIP-based terminal mobility [90] takes care of binding update by application layer signaling. It reducesone-way packet delay by avoiding the triangular routing that is inherently present in Mobile IPv4. I aug-ment the SIP-based terminal mobility with a complete set of hand-off operations to support inter-domainmobility [25]. The augmented prototype includes a combination of network detection, registration, con-figuration, authentication, security, location management functions, and roaming support for the wirelessInternet [25]. I compare the performance of SIP-based mobility with MIPv4 for both subnet and inter-domain mobility and verify the effect of route optimization.
Initially, I compared the latency due to SIP and MIP for different packet sizes during subnet hand-off. By using SIP-based mobility protocol I could obtain 50 percent latency improvement in real-time(RTP/UDP) traffic thus providing a reduction in latency from a baseline of 27 ms to 16 ms for largepackets and a 35 percent utilization increase by having 60 bytes/packet compared with baseline of 80bytes/packet size with IP-in-IP encapsulation in Mobile IP for a payload of 40 bytes. These resultsdemonstrate how a direct signaling path between CH (Correspondent Host) and MH (Mobile Host) op-timizes end-to-end delay and reduces data overhead.
In order to validate the effect of route optimization and study the effect of security and authenticationon the handoff process, I experimented with secured inter-domain mobility and compared both networklayer (e.g., Mobile IP) and application layer (e.g. SIP-based) mobility techniques [91] and [92]. Securedinter-domain mobility introduces additional operations, such as local authentication and interaction withan authorization server resulting in additional delay.
Table 8 shows the timing delay for some of the functional components during inter-domain handoffusing both SIP and MIP as the mobility protocols. End systems processing time are not shown here.
19
These experimental results demonstrate delays due to several handoff components, such as layer 2 bea-con interval, subnet, and domain discovery, IP address acquisition, local authentication, authorization,and delay due to binding update, such as SIP re-INVITE and MIP registration. RTP1 represents the timewhen the RTP media is received prior to handoff in the old network and RTP2 represents the time RTPmedia is received in the new point of attachment after the handoff. For both the cases, IKE (Internet KeyExchange) took almost 5 seconds for successful establishment of the IPSEC-based security association.
Table 8: Inter-domain handoff results
MobilityType
RTP1Media
Layer 2BeaconAdv.
RouterAdvertisement
Layer 3Configuration
AuthenticationDelay
IKEProcess
BindingUpdate
RTP2Media
In-handoffDelay
SIP 51.7 s 120 ms 500 ms 80 ms 10 ms 5130ms
240 ms 53 s 1.3 s
MIP 23.8 s 120 ms 500 ms 80 ms 10 ms 4580ms
20 ms 31.1 s 7.3 s
These experimental results also verify the fact that by maintaining the direct path between MH andCH one can optimize inter-domain mobility.
6.5.2 Interceptor-assisted packet modifier
I have augmented the original MIP-LR scheme by Jain et al. [5] with application layer modules, suchas packet interceptor and packet modifier. The packet interceptor and modifier modules at both senderand receiver side cooperate with each other to provide route optimization and tunnel optimization. Inter-ceptor modules intercept the outgoing packets and mangler modules change the destination address ofthe packet at the sender side before it is transmitted. The mangler module at the receiver side changesthe destination address back to the permanent address of the mobile before sending it to the application.This way, the application is not aware of the underlying IP address change and the packets from thecorrespondent host are sent directly to the mobile’s new point of attachment. From the experimentalresults in a testbed, I have verified that [93] one can attain up to 50 percent reduction in managementoverhead and up to 40 percent improvement on latency compared to standard mobile IP in co-locatedmode. Table 9 shows round trip time delay comparison between the interceptor assisted MIP-LR andMobile IP. These results verify the effect of proposed techniques by providing direct path between CHand MH.
Table 9: Experimental validation of route optimization
PacketSizeBytes
One Way DelayEmulated delay0
One Way DelayEmulated delay(10 ms)
One Way DelayEmulated delay(20 ms)
One Way DelayEmulated delay(40 ms)
MIP(RTT)(ms)
MIP-LR(RTT)(ms)
MIP(RTT)(ms)
MIP-LR(RTT)(ms)
MIP(RTT)(ms)
MIP-LR(RTT)(ms)
MIP(RTT)(ms)
MIP-LR(RTT)(ms)
64 5.9 4.1 25.3 5.4 45.2 5.9 85.3 7.2128 6.6 4.7 27.4 5.6 46 6.9 86.3 8.6256 8 5.9 28.1 6.2 48 7.8 88.2 10512 13.9 10.2 32 10.8 51.9 11.4 92 13.81024 19.5 13 39.5 14.2 59.4 15.4 99.7 16.9
6.5.3 Midcom proxy-assisted route optimization
Forwarding of application layer signaling is delayed due to the routing indirection caused by the under-lying network layer mobility mechanism. This gives rise to additional traversal path for the application
20
layer signaling even if the destination node is closer the mobile. As an example, in MIPv4, reverse tun-neling at FA (Foreign Agent) forces the packet from a mobile node to route via the home agent givingrise to the additional route traversal. At the application layer, SIP signaling is usually routed via theoutbound proxy (e.g., SIP server) that is closer the mobile. However, due to the indirection imposed bythe underlying mobility layer, these packets need to travel to the home network before being directed tothe application servers in the edge of the network. This additional routing causes signaling delay andaffects the handoff.
I propose a midcom proxy-assisted technique to optimize the packet delivery between MN (MobileNode) and local resources, such as outbound SIP server. The proposed mechanism introduces packetinterceptors and forwarding modules both in FA and the local SIP server to perform selective tunnelingoperation in both directions. Packets received at the FA via the server-FA tunnel get decapsulated at theFA and forwarded to the MN. In addition, it does not require any changes in the existing MIP modules.I applied this technique to optimize the path for SIP signaling messages, such as re-REGISTER andre-INVITE in a MIPv4-based mobility environment. Using this technique, I demonstrated that the ap-plication layer signaling packets are not affected by the underlying indirection imposed by the mobilityprotocol. For example, applying this technique, SIP re-registration took about 3205 ms compared to4025 ms in the absence of this optimization technique. Similarly, SIP Re-INVITE took about 1660 mscompared to 2502 ms in the absence of mitigation technique. I describe the details of the results in [94].Effect of this mitigation technique is more pronounced when the server is in the visited domain, closerto the mobile node,and the home network is far from the visited network.
6.6 Binding updateDistance between the mobile node and correspondent node adds to the binding update delay resulting inoverall handoff delay and data loss. In this section, I propose several optimization techniques to optimizethe binding update delay and reduce the effect of binding update delay.
6.6.1 Hierarchical binding update
Enhancements to layer 3-based mobility protocols reduce the binding update when CN (CorrespondentNode) and MN are far apart. HMIP [95] provides hierarchical mobile IP registration for IPv4. HMIPv6[96] introduces an agent called MAP (Mobility Anchor Point) to localize the intra-domain mobility man-agement. I have developed and demonstrated hierarchical binding update techniques for both applicationlayer and network layer mobility protocols. Details of the implementation and experimental analysis forthe above two cases are described in [97] and [9], respectively. I describe these two techniques briefly.
The proposed techniques introduce an anchor point in the network that helps to limit the bindingupdate when the mobile’s movement is confined to a domain. I have applied these techniques to boththe application layer mobility and network layer mobility protocol.
For the case of an application layer mobility protocol, I use a back-to-back SIP user agent (B2BUA)as the anchor point, possibly closer to the mobile node. A B2BUA consists of two SIP user agents whereone user agent receives a SIP request, possibly transforms it and then has the other part of the B2BUAre-issue the request. A B2BUA in each domain needs to be addressed by the MH in the visited domain.The B2BUA issues a new request to the CH containing its own address as the media destination and thenforwards the packets, via RTP translation or NAT, to the MH.
I have designed a network layer intra-domain mobility management [9] protocol by adopting a sim-ilar approach where a mobility agent acts like an anchor point. The mobile assigns two addresses: localcare-of-address and global-care-of-address. The first time the mobile moves to a domain, it sends twobinding updates, one to the mobility agent with local care-of-address and one to its home agent with theaddress of the mobility agent which is same as the global care-of address. Thus, any packet from thehome agent gets intercepted by the local mobility agent first. The local mobility agent first decapsulatesthe original packet, then encapsulates it again with local-care-of-address and then sends it to the mobile.For every subsequent move within the domain, the local binding update is sent to the anchor agent only
21
and is not propagated to the home agent. Although this technique reduces the delay to binding update,the mobile may suffer from additional delay due to encapsulation and decapsulation at the mobilityagent.
6.6.2 Proactive binding update
Proactive binding updates allow the mobile to send a binding update before the mobile has moved tothe new network. It helps to eliminate the delay due to binding update after the handoff. FMIPv6 [98]adopts a Fast Binding Update (FBU) technique where it sends the binding update to the access routersso that the in-flight packets during handoff can be forwarded from the previous access router to themobile. However, this requires a signaling handshake between the neighboring routers to forward thedata. Malki et al. [99] also describe techniques to provide low latency handoff in MIPv4 environmentwhere the transient packets are forwarded from the previous foreign agent.
I have designed a proactive binding update technique that works in conjunction with the preauthen-tication and proactive configuration techniques described earlier in this section. As part of the proposedtechnique, the mobile uses the IP-IP tunnel and sends the binding update proactively with the cachedIP address that it has obtained from the new network. Thus, any packet destined to the new addressis picked up by the next access router and is tunneled to the mobile before it moves to the new pointof attachment. After the mobile moves to the new network, the IP-IP tunnel is deleted and the new IPaddress is assigned to the physical interface, but no new binding update is necessary [71]. This effec-tively removes the need for sending another binding update after the mobile moves to the new network.Thus, it reduces the delay associated with the binding update altogether. This technique uses the generaloptimization principle of reducing the amount of signaling messages after the handoff in establishing thenew identifier. However, it introduces the additional complexity of managing the transient tunnel.
6.7 Redirection of in-flight dataRedirecting the transient media during handoff reduces the in-handoff packet loss. There is a smallamount of related work [99], [100], and [101] that help reduce the transient data loss during handoff byredirecting the data from the previous network. Vakil et al. [102] provide a virtual soft-handoff approachfor CDMA-based wireless IP networks. However, this scheme works for CDMA network only and doesnot provide a generalized solution suitable for other type of access network, such as 802.11.
In this section, I describe two redirection techniques that I have developed. These are mobilityproxy-assisted time-bound data redirection and timebound multicasting. These techniques help mitigatethe effect of binding update delay by forwarding the in-flight data to the new point of attachment. Thesetechniques were developed at the same time as the related work. Compared to the related work, theproposed techniques do not need any changes in the existing networking infrastructure.
6.7.1 Mobility-proxy assisted time-bound data redirection
First technique involves the general principle of packet interception and forwarding that uses a mobility-proxy capturing the in-flight data in the previous network and forwarding it to the new destination of themobile. I have implemented this technique in a SIP-based environment. In case of intra-domain mobility,each visited domain may consist of several subnets. For SIP-based mobility, every move to a new subnetwithin a domain causes the MH (Mobile Host) to send a re-INVITE to the CH containing its new care-ofaddress. If the re-INVITE request gets delayed due to path length or congestion, transient media packetswill continue to be directed to the old address and thus get lost. These proposed techniques help alleviatetransient data loss resulting out of continuous hand-offs within a domain and thus minimize the effectof delay contributed during application layer rebinding. In-flight packets can be redirected to a unicastor multicast address based on the movement pattern of the mobiles and usage scenario. I experimentwith SIP registrar and RTP translator or NAT, the outbound proxy, and a mobility proxy to implement
22
these mechanisms. I provide the details about the fast-handoff mechanism in [103]. I briefly describetwo different ways of how I have applied these two techniques in the following paragraphs.
In the experimental testbed the visited network has an outbound proxy. I enhance this proxy withthe ability to temporarily register visitors [104]. The visitor obtains a temporary, random identity fromthe visited network and uses it as its new address-of-record to register with the registrar in the visitednetwork. The hierarchical registration speeds up the registration, but does not address the “delayedbinding update” issue using SIP’s re-INVITE feature if the CH is very far. I have taken care of thedelayed binding update using a mobility proxy assisted technique.
In the experiment, each subnet within a domain is equipped with a mobility proxy that has the abilityto intercept the packet destined to mobile’s old address and forward it to the new destination. RTPtranslator [105] provides an application-layer forwarding technique that can forward the RTP packetsfor a given address and UDP port to a given network destination. SIP requests typically traverse a SIPproxy in the visited network, the outbound proxy. As the mobile moves to a new network, it sendsboth re-REGISTER and re-INVITE messages via the outbound proxy. This outbound proxy can beconfigured as visited registrar. Thus, the visited-network registrar receives the registration updates fromthe MH that has just moved, and immediately sends a request to the mobility proxy in the network thatthe MH just left. The request causes the mobility proxy to intercept the packets and the RTP translatorforwards any incoming packets to the new address of the MH. After a set interval or after no mediapackets are received by the RTP translator, the mobility proxy relinquishes this old address and removesthe forwarding table entry, assuming that the re-INVITE has reached the CH.
Alternatively, the outbound proxy can use the data in the MH-to-CH re-INVITE to configure themobility proxy in the previous network. The advantage of this approach is that the outbound proxyusually has access to the Session Description Protocol (SDP) information containing the MH mediaaddress and port, thus simplifying the configuration of the translator or NAT. On the other hand, thisoutbound proxy has to remember the INVITE information for an unbounded amount of time and becomecall stateful, since it needs the old information when a new re-INVITE is issued by the MH. Usingmobility proxy-based approach, I have verified this technique using two different tools, such as rtptrans[106] and Linux iptables [107] that help direct the transient traffic from the previous subnet to the newone.
In the experimental testbed, RTP translators are associated with the mobility proxies in the respectivesubnets. Mobility proxy in each of these subnets intercepts the traffic meant for the mobile host andRTP translator forwards it to the new address of the mobile host after capturing it. This is achieved by acombination of SIP-CGI and SIP REGISTER [108]. Signal re-INVITE was delayed to simulate distancebetween CH and MH. Both VIC and RAT tools [109] were used to measure the performance of audioand video streaming traffic, respectively. I delayed the traversal of re-INVITE signals by 100 ms, 200ms, 500 ms, 1 sec, 2 sec and 3 sec to show how RTP translator helps the media redirection and mitigatesthe packet loss during mobile’s movement. I measured the packet forwarding delay due to redirectionat the registrar to be approximately 1 ms when the iptables-based NAT approach was used, whereas theRTPtrans approach added 4 ms of delay. The additional delay is due to application layer redirection usedby RTPtrans.
6.7.2 Small group multicasting
Locally scoped multicast technique allows to multicast transient data to the neighboring networks. Ithelps to avoid packet loss if the MH can predict that it is about to move to a new subnet within theneighboring network. I have implemented this mechanism for both application layer and network layermobility protocols. In case of application layer mobility, the mobile informs the visited registrar orB2BUA of a temporary multicast address as its contact address in the SDP. Once the MH has arrivedin its new subnet, it updates the registrar or B2BUA with its new unicast address, while it continuesto receive the transient data over the multicast address. In case of network layer mobility, the mobilityagent (MA) encapsulates the unicast packets in a multicast address and sends it to the neighboring basestations proactively. The base stations buffer these packets and upon the arrival of the mobile, the unicast
23
packets are delivered to the client. Details of media redirection for network layer mobility protocol canbe found in [110]. The use of locally scoped multicast is only effective if the MH can quickly acquirea multicast address and there is a multicast infrastructure available. It is also subjected to additionalencapsulation.
Thus, using the above two techniques, the effect of binding update delay is minimized by redirectionof the in-flight transient packets.
6.8 Network bufferingAlthough media redirection techniques help redirect the in-flight packets caused due to delayed bindingupdate, some packets may be lost during link-layer handover. Thus, it is essential to mitigate the effectof handover delay by reducing the pasket loss.
Bicasting or buffering the transient packets at the access routers can be applied to minimize oreliminate the packet loss. However, bicasting alone cannot eliminate packet loss if link-layer handoveris not seamless. In addition, buffering in the network node introduces an additional one-way delayfor the in-flight packets, even if it mitigates the effect of handoff by reducing the packet loss. Whilethis additional end-to-end delay may not affect the streaming traffic, interactive traffic, such as VoIPapplication cannot tolerate the jitter resulting out of larger one-way delay. Moore et al. [111] andKrishnamurthi et al. [112] have developed buffering techniques for MIPv4 and MIPv6, respectively.Most multimedia applications resort to FEC [113], RTCP-based feedback [114] and other stream repairtechniques [115] to reduce the effect of packet loss in a lossy wireless medium.
However, none of these mechanisms has been applied to take care of packet loss during handoff.I have designed and implemented a dynamic buffering scheme [116] that ensures zero packet loss atthe expense of additional end-to-end delay that can be controlled. The buffering scheme is used inconjunction with the existing mobility protocols, access protocols or as an independent network or linklayer mechanism. Ability to control the buffer dynamically provides a reasonable trade-off betweendelay and packet loss within the threshold limit for real-time communication. I have experimented withtwo kinds of buffering scheme, such as time limited buffering and explicit signaling buffering. In caseof time limited buffering technique, the mobile and buffering node can negotiate a buffering period thatis conveyed to the buffering node during the initial setup signal. The buffering node actually buffersthe packets for the duration of the buffering period as defined in the initial control message. In caseof explicit signal buffering, the buffering period is equivalent to the total handoff period and additionaltime taken to flush the buffer after the handoff has taken place. I have experimented with both typesof buffering schemes using different traffic rates and buffering periods. Average packet delay, packetloss and average number of packets buffered were calculated for each case. In case of explicit signaling,the number of packets buffered were dependent upon handoff delay and packet generation rate. Timelimited buffering on the other hand introduces higher probability of losing the packets in case of bufferoverflow.
6.9 Cross-layer triggersCross layer triggers are useful hints that can expedite the series of sequential handoff operations. Thesecould be applied to several stages in the handoff process, such as discovery, configuration and acrosslayers. Lower layer events are generally passed as the triggers to upper layers so that these mobility re-lated functions can be expedited. The lower layer triggers can also prepare the mobile for the impendinghandover event by preparing different phases of the handover proactively. The layer 2 triggers can assistlayer 3 operations that rely exclusively on these indicators to perform specific actions such as detectionof network attachment or detachment resulting in better handoff operations.
I have designed cross layer triggers that assist the mobile during impending movement or during sud-den disconnection. Many of these techniques are being standardized as “Command Service” and “EventService” within IEEE 802.21. Optimizing some of the event triggers, such as “link down detection” and“link up detection” techniques are useful to optimize the overall handoff operations.
24
In 802.11-based layer 2 operation, a client that is currently associated with an AP (Accesss Point)may experience sudden disconnection due to device failure in the AP or in the client or perhaps an un-anticipated rapid movement of the client that quickly brings it out of range of the AP. Using signal qualityto immediately determine “link down” events in the client during these scenarios can be misleading sincea client registers the link quality based only on the last received frames. Therefore, link quality only rep-resents historical data and it is reset only after certain number of expected beacon frames have not beenreceived. Other implementations verify connectivity using failed transmission events (RTS/CTS) [63]. Ihave designed an optimized link detection technique that uses a combined scheme of passive monitoringof 802.11 frames as well as active probing of the AP at some defined conditions. A combination of theseschemes as well as monitoring of independent indicators provides several link-down detection variantsthat is applicable to different scenarios. By this method, the mobile can rapidly determine the suddendisconnection event and quickly propagates “link down” indications to upper layer protocols, such asmobile IP that maybe present in the mobile.
6.10 Cross layer mobility optimizationIn this section, I describe a cross layer mobility optimization technique that uses triggers from MAClayer and application layer and optimizes several handoff operations, such as address configuration,layer 3 binding update, and media traversal. During the discovery process, MAC layer trigger specifiesthe mobile’s movement type based on layer 2 beacon id and layer 3 subnet or domain id. Similarly,application layer trigger specifies the type of application such as TCP-based (e.g., ftp) or RTP-based(e.g., VoIP) on the mobile. An equivalent Petri net model of this mechanism can be derived based onFigure 3 in Section 5. This mechanism optimizes the overall system performance by way of limitingthe number of signaling update message during a mobile’s movement within a domain and reducing themedia traversal by use of direct binding update for real-time application.
During intra-domain movement, a local mobility protocol takes care of redirecting the traffic to thenew point-of-attachment of the mobile and confines the binding update to the domain itself. Mobilityprotocols at each layer are best suited to work for a specific type of application (e.g., TCP-based andRTP-based) and movement pattern (e.g., inter-domain and inra-domain) of the mobile. For example,an application layer mobility protocol may be suited to work for interactive traffic such as VoIP and anetwork layer mobility protocol works well to support TCP-based application. Cross layer optimizationtechniques help the mobile to reduce the handoff delay by avoiding the layer 3 configuration, limitingthe binding update and reducing the media traversal by choosing the appropriate mobility protocol.
I have designed, prototyped and validated this cross layer mobility optimization scheme using threemobility protocols such as SIP-based [90], MIP-LR [93] and MMP (Micro Mobility Protocol) [117] thatoperate at application layer, network layer and at layer 2, respectively. These mobility protocols aretriggered by the cross layer triggers and perform the desired optimized operations.
MMP is a modified version of Cellular IP [118] that provides additional survivable features in anad hoc network. SIP-based mobility protocol is used for real-time traffic, and application layer MIP-LRis used for non-real-time traffic during a node’s movement between two different domains while MMPtakes care of the movement within a domain. MMP is designed as a micro-mobility protocol to handleintra-domain mobility and works in conjunction with SIP and MIP-LR, that handle macro-mobilityand take care of inter-domain movement. To support real-time communication during the mobile’smovement between the domains, the mobile sends a re-INVITE to the CH to keep the session active,similarly a MIP-LR UPDATE message is sent to CH for the TCP/IP traffic. But for any subsequentmove within the new domain, re-INVITE or update messages are not sent, since MMP takes care ofrouting the packets properly within that domain. This helps limiting the binding update from traversinga long distance.
Figure 6 illustrates a scenario where macro-mobility and micro-mobility protocols work togetherbased on MAC layer and application layer triggers. The results shown in Figure 6 demonstrate how themobile can obtain higher throughput when a micromobility protocol is used for intra-domain movementcompared to a macro mobility protocol because it limits the number of signaling message update. I
25
1 2 3 4
gateway
BS’sMH
1 2 3 4
gateway1
BS’sMH
Host BasedRoutingMicro-MobilityDomain
Host BasedRoutingMicro-MobilityDomain(Cellular IP)
CH
Macro Mobility (SIP, MIP-LR)
gateway2Micro mobilityDomain 1
Micro mobilityDomain 2
5 6 05
1015202530
0 0.05 0.1 0.15 0.2 0.25 0.3Link Latency (s)
Thro
ughp
ut (k
byte
/s) MIP 1
MIP 0.2MIP 0.1HBR SS 1HBR SS 0.2HBR SS 0.1
Figure 6: Cross layer mobility optimization scenario and results
describe the complete implementation of the cross layer mobility management scheme in [119]. Simula-tion and experimental results for MMP are explained in [117]. I have described the architectural detailsof this scheme in [120]. Most recently, the IETF is considering a network controlled localized mobilityprotocol [121] that adopts a similar optimization technique and helps to optimize handoff during a mo-bile’s movement within a domain. It introduces a new network element, such as LMA (Local MobilityAgent) that separates the mobility domains. Based on mobile’s movement, it limits the binding updatewithin LMA domain.
6.11 Systems evaluationIn this section, I describe the experimental results from a mobility system that uses a set of proactiveoptimization techniques that I have described earlier to support multiple types of handoff such as intra-technology and inter-technology.
I have prototyped a mobility system called media independent pre-authentication [88] that utilizesmany of the techniques that I have developed to optimize the basic operations, such as network discovery,authentication, configuration, security association, and binding update. This system also optimizes thelink layer handoff delay by avoiding scanning and applies the cross layer optimization techniques, suchas fast “link down” detection. In addition, it reduces packet loss by implementing the dynamic bufferingand copy-and-forward mechanism [116] at the edges of the network. I have applied these proactivetechniques to optimize both network layer and application layer mobility protocols over both singleinterface or multiple interfaces [26], [88].
FMIPv6 [98] and Gwon et al. [122] have developed a few mobility systems that utilize the proac-tive handoff techniques for both MIPv6 and MIPv4, respectively. However, these systems do not ad-dress proactive discovery or preauthentication mechanisms. Also, these mechanisms require signalingexchange between the neighboring routers that works only if the routers have trust relationship. I per-formed a comparative analysis of Media Independent Pre-authentication system with similar fast-handoffapproaches, such as proactive mode of FMIPv6 and have described the details of the results in [123].
6.11.1 Intra technology handoff
Table 10 shows the experimental results from a mobility system that demonstrate intra technology han-dover involving 802.11. Both application layer mobility such as SIP and network layer mobility such asMIPv6 were used for experimental validation of these techniques. These results demonstrate how sev-eral of the proactive optimization techniques can work together to minimize the delay and packet loss,
26
jitter and delay. Results also demonstrate the effect of buffering in reducing packet loss at the expenseof added delay.
Table 10: Delay and packet loss during proactive handoff
Mobility Type MIPv6 SIP MobilityHandoffoptimizationParameters
BufferingDisabled+RODisabled
BufferingEnabled+RODisabled
BufferingDisabled+ROEnabled
BufferingEnabled+ROEnabled
BufferingDisabled
BufferingEnabled
L2 Handoff(ms)
4.00 4.3 4.00 4.00 4.00 5.00
Avg. packet loss 1.3 0 0.7 0 1.50 0Avg. Inter-packetinterval (ms)
16.00 16.00 16.00 16.00 16.00 16.00
Avg. Inter-packetarrival duringhandover (ms)
21 45 21 67 21 29.00
Avg. packet jitter(ms)
n/a 29.3 n/a 50.6 n/a 13.00
Buffering period(ms)
n/a 50.00 n/a 50.00 n/a 20.00
Avg. BufferedPackets
n/a 2.00 n/a 3.00 n/a 3.00
6.11.2 Inter technology handoff
Inter technology handoff involves handoff between multiple interfaces each belonging to different ac-cess technology. I have applied the optimization techniques to inter-technology handover for severalscenarios.
In one scenario, the new interface comes up only after the link to the old interface is down. Thisscenario usually gives rise to undesirable packet loss and handoff delay and can be termed as “break-before-make”. Lower layer triggers, such as “Link Down” could be useful here. In the second scenario,the second interface is prepared proactively while the mobile still communicates using the old interface,and at some point the mobile decides to use the second interface as the active interface and completesthe authentication and binding update procedures. This results in fewer packet loss as it uses make-before-break techniques to set up the layer 2 configuration. In the third scenario, some of the requiredfunctions, such as nework selection, context transfer of CDMA network parameters such as PPP stateand security associations are established ahead of time using the current interface. This scenario maybe beneficial from a battery management standpoint. By activating the second interface only after anappropriate network has been selected and the mobile has authenticated itself using the old interface itmay utilize battery more efficiently.
In order to optimize scenario 1, I have applied the fast “Link Down” detection technique along withpre-authentication, proactive configuration techniques, buffering, and copy-and-forwarding technique toreduce the packet loss and delay. The mobile was able to resume the communication in the new network,within 50 ms and the packet loss was reduced to 0.
In a typical break-before-make handoff without any optimization, the mobile prepares the CDMAinterface only after current interface (802.11) went down. The handoff delay and associated packetloss resulted because of PPP configuration delay (16 sec) and binding update delay, such as SIP re-INVITE (1.5 s) and MIP registration (500 ms). As part of scenario 2, I have developed the make-before-break algorithm to support handoff between CDMA and 802.11 networks. Using this technique,the mobile sets up the layer 3 configuration in the CDMA network using the PPP interface while it
27
still keeps on communicating using the current 802.11 interface. This technique reduces the packetloss due to delay in layer 2 and layer 3 configuration. I have experimented with this mechanism forboth network layer protocol, such as MIPv4 and application layer protocol such as SIP-based mobilityover 802.11 and CDMA1XRTT. By using make-before-break technique, there was no packet loss forboth MIP and SIP. However, initial jitter was observed for the in-flight packets after the handover from802.11 to CDMA network and out-of-order packets were received after the handover from CDMA to802.11 networks. I have published a complete experimental handoff analysis of this work in [26]. Fromthe experiments it is verified that make-before-break technique aided by a combination of proactiveoptimization methodologies and cross layer triggers can reduce the delay during handover.
7 Roadmap for Future WorkIn this section, I provide a roadmap for the completion of my research and describe briefly the additionalwork that I will perform beyond my proposal defense.
7.1 Verification of optimal handoff operation using Petri net modelsI will apply the proposed optimization techniques to Petri net-based mobility model and will completethe performance evaluation of the optimized models using TimeNet [124] and Floyd [59] algorithm. Petrinet-based optimization techniques for the mobility event can primarily be mapped into three categories- sequential, concurrent, and proactive. Network bandwidth, memory, processing power are some ofthe resources that are utilized during any mobility event. Depending upon the type of optimization,amounts of systems resources expensed during a specific operation will vary over a period of time.While a sequential handoff operation takes more time compared to a proactive or concurrent operation,the optimized models using concurrent or proactive operations may need to exhaust more resources fora given period of time. In Figure 7, I give an example of how a specific handoff operation (e.g., layer 2discovery) can be represented in Petri net model using sequential, concurrent and proactive optimizationtechniques. Number of tokens represents amount of resources exhausted during each of these operations.For example, one token can represent 100 kb of bandwidth. I will investigate the optimality of the systemperformance by comparing the handoff performance (cycle time) and resource utilization (number oftokens) for these handoff methodologies. Expected completion date is May 2008.
7.2 Analysis of optimal proactive handoffWhile the proactive operations reduce the delay and packet loss, these also utilize additional resourcesof the mobile and different networking components. In order to reduce the effect of failed handover,mobile may often need to complete proactive operations with multiple neighboring networks. This mayresult in additional resource utilization in the network. These additional resources could include the costfor discovery of neighboring network parameters, cost for tunnel setup, cost for keeping the IP addressin the cache, cost for sending multiple stream of data via the multiple tunnels, cost associated withadditional bandwidth due to flow via the transient tunnels, and tunnel management cost. Based on theextent of proactive handover operations, such as preauthentication, preconfiguration or proactive bindingupdate, associated resource utilization and cost vary. As part of the future work, I will analyze optimalhandoff probability at the cost of additional resources for various proactive handover operations. I willuse Petri net model and simulated annealing techniques to perform this cost-benefit analysis. Expectedcompletion date is August 2008.
7.3 Tradeoff analysis between in-flight packet loss and one-way delayWhile buffering mechanism in the network reduces the packet loss, it also introduces additional delayto the one-way-delay of any packet and contributes to the jitter for in-flight packets that are buffered.An in-flight packet that would have got lost otherwise gets buffered in the buffering node for a certain
28
ResourcesNetwork b/w
ResourcesMemory
ResourcesNetwork b/w
ResourcesPower
scanning Open/Authentication
4-way handshake
t1 t2 t3 t4
P1 P2 P3
2
Association
Connected Disconnected
1
a. Sequential Operation
ResourcesNetwork b/w
Resourcesprocessingpower
ResourcesPower
4-way handshakecomplete
t2
t3
t4
P2
P3
t1Scanningoperation
Open/Authentication
Scanningcomplete
4-wayHandshakeOperation
P1
ResourcesNetwork b/w
1
2
Authenticated
1 token = 100 kb2 tokens = 200 kb
ConnectedAssociation
b. Concurrent Operation
Resources(Network b/w)
Resources(Memory)
scanning
P11
t12
ResourcesNetwork b/w
Power
Open/Authentication
4-way handshake
t2 t3 t4P2 P3
AssociationConnected
P12
Disconnected
caching
c. Proactive Operation
Figure 7: Resource utilization for sequential, concurrent and proactive operations
period of time. This time is determined by the handoff delay and time taken to flush the packets from thebuffer at the new point of attachment. Packets arrive in the buffer at a regular interval. But when thesepackets are flushed out of the buffer after the handoff, these buffered packets do not maintain the inter-packet arrival gap resulting in jitter. Thus, although buffering mechanism avoids packet loss, the mobileis subjected to a spike in traffic since the buffered packets arrive on the mobile almost instantaneously.
The in-flight packets that are buffered in the edge router are subjected to an increased amount of delaycompared to pre-handoff and post-handoff packets. But each consecutive in-flight packet is subjected toa different amount of one-way-delay since these packets spend different amount of time in the buffer.Later packets are subjected to lesser amount of delay compared to the packets that were queued in first.Figure 8 shows how the in-flight packets are affected due to the buffering operation at the edge routerof the network. End-to-end delay for in-flight packets, delay between last packet in the old networkand first packet in the new network, and total number of packets that are affected due to buffer areimportant metrics that should be considered for real-time communication. Packet generation rate at thesource, handoff period, time taken to signal the buffer to flush, packet transmission time are some ofthe parameters that affect the optimal buffer length at the router. I will complete the tradeoff analysisbetween packet loss and one-way-delay of the packets due to end system buffering during handoff. Thiswill analyze how placement of buffer, handoff delay, and buffer length may affect the amount of jitterfor the in-flight packets. Expected completion date is October 2008.
29
Buffering Node
Pre Handoff Traffic
Buffered traffic
New Network Previous Network
Mobile Node
Correspondent Node
Flushed Traffic after handoff
Post Handoff Traffic
Signaling
Access Router (BN)
Access Router
a. Buffering at the Network Edges b. Effect of buffering on in-handoff packets
0
20
40
60
80
100
120
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Packet Number
One
way
pac
ket d
elay
(ms)
End-to-end packetdelay due to buffering
Pre-handoffpackets
Pre-handoffpackets
Post-handoffpackets
Figure 8: Effect of network buffering on in-flight packets
7.4 Summary of plan for completion of researchTable 11 shows my plan for completion of the research. Thus, I plan to defend my thesis in February
Timeline Work ProgressMobility systems analysis completedPreliminary mobility systems modeling completedMobility optimization methodologies completedPrototyping and experimental verification of mobility optimization methodologies completed
May. 2008 Verification of optimal handoff operation using Petri net models ongoingAug. 2008 Analysis of optimal proactive handoff ongoingOct. 2008 Tradeoff analysis between in-flight packet loss and one-way delay ongoingNov. 2008 Thesis writingFeb. 2009 Thesis defense
Table 11: Plan for completion of my research
2009.
8 ConclusionThis thesis proposal contributes to the general theory of optimized handover and has addressed the needfor a formal systems model that can characterize a mobility event and the associated formal mobilityoptimization methodologies. It provides a systematic and formal approach to a mobility event. Aftera thorough analysis of the abstract operations associated with several mobility protocols, I determinedthat these basic operations form a set of discrete events that can be modeled as Discrete Event DynamicSystem. I used Deterministic Timed Petri net to model the mobility event and conducted the performanceanalysis. This model can characterize certain optimization methodologies leading to a set of designprinciples for any new mobility protocol as well as evaluate its effectiveness. Initial results includeidentification of basic properties of a mobility event, formulation of a formal systems mobility model,design of few optimization methodologies based on some fundamental design principles of optimizationand evaluation of the associated optimization techniques by means of analysis, model-based simulation,and experiments for different mobility functions. Roadmap and timeline to completion of the researchhave been laid out. These include performance analysis using Petri net model, tradeoff analysis ofoptimization techniques and resource utilization, analysis of optimal proactive handoff, and tradeoffanalysis between delay and packet loss.
30
References[1] N. Tripathi, J. Reed, and H. F. VanLandingham, “Handoff in cellular systems,” IEEE Personal
Communications Magazine, vol. 5, Dec. 1998.
[2] C. Perkins, “IP mobility support for IPv4,” RFC 3344, Internet Engineering Task Force, Aug.2002.
[3] D. Johnson, C. Perkins, and J. Arkko, “Mobility support in IPv6,” RFC 3775, Internet EngineeringTask Force, June 2004.
[4] C. Perkins, “Mobile IP,” Communications Magazine, IEEE, vol. 40, no. 5, pp. 66–82, 2002.
[5] R. Jain, T. Raleigh, D. Yang, L. F. Chang, C. J. Graff, M. Bereschinsky, and M. Patel, “Enhancingsurvivability of mobile Internet access using mobile IP with location registers,” in Proceedings ofthe Conference on Computer Communications (IEEE Infocom), (New York), Mar. 1999.
[6] E. S. Eronen, “IKEv2 mobility and multihoming protocol,” RFC 4555, Internet Engineering TaskForce, 2006.
[7] A. T. Campbell, J. Gomez, S. Kim, A. G. Valk, C.-Y. Wan, and Z. R. Turnyi, “Design, implementa-tion, and evaluation of cellular IP,” IEEE Personal Communications Magazine, vol. 7, pp. 42–49,Aug. 2000.
[8] R. Ramjee, T. F. LaPorta, L. Salgarelli, S. Thuel, K. Varadhan, and L. E. Li, “IP-based accessnetwork infrastructure for next-generation wireless networks,” IEEE Personal CommunicationsMagazine, vol. 7, pp. 34–41, Aug. 2000.
[9] S. Das, A. Dutta, A. McAuley, A. Misra, and S. Das, “IDMP: an intra-domain mobility manage-ment protocol for next generation, wireless networks,” IEEE Personal Communications Maga-zine, June 2002.
[10] D. A. Maltz and P. Bhagwat, “MSOCKS: an architecture for transport layer mobility,” in Pro-ceedings of the Conference on Computer Communications (IEEE Infocom), (San Francisco, Cal-ifornia), p. 1037, March/April 1998.
[11] A. C. Snoeren and H. Balakrishnan, “An end-to-end approach to host mobility,” in ACM/IEEEInternational Conference on Mobile Computing and Networking (MobiCom), (Boston, Mas-sachusetts, USA), pp. 155–166, Aug. 2000.
[12] S. Koh et al., “Use of SCTP for seamless handover,” internet draft, Internet Engineering TaskForce, Feb. 2003. Work in progress.
[13] H. Schulzrinne and E. Wedlund, “Application-layer mobility using SIP,” Mobile Computing andCommunications Review (MC2R), vol. 4, pp. 47–57, July 2000.
[14] J. Rosenberg, H. Schulzrinne, G. Camarillo, A. Johnston, J. Peterson, R. Sparks, M. Handley, andE. Schooler, “SIP: session initiation protocol,” RFC 3261, Internet Engineering Task Force, June2002.
[15] R. Moskowitz, P. Nikander, P. Jokela, and T. Henderson, “Host identity protocol,” Internet Draftdraft-ietf-hip-base-10, Internet Engineering Task Force, Oct. 2007. Work in progress.
[16] G. Xylomenos and G. C. Polyzos, “IP multicast for mobile hosts,” IEEE Communications Maga-zine, vol. 35, Jan. 1997.
31
[17] U. Varshney and S. Chatterji, “Architectural issues in IP multicasting over wireless networks,” inIEEE conference on Wireless Communication and Networking Conference, (New Orleans), IEEE,June 1999.
[18] A. Acharya, A. Bakre, and B. Nath, “IP multicast extensions for mobile internetworking,”Technical Report LCSR-TR-243, Department of Computer Science, Rutgers University, NewBrunswick, New Jersey, Apr. 1995.
[19] I. Romdhani, M. Kellil, H.-Y. Lach, A. Bouabdallah, and H. Bettaher, “IP mobile multicast:Challenges, solutions and open issues,” Motorola technical report, Motorola Labs, Paris, France,July 2002.
[20] C. Williamson, T. Harrison, W. L. Mackrell, and R. B. Bunt, “Performance evaluation of the MoMmobile multicast protocol,” ACM Mobile Networks and Applications (MONET) Journal, vol. 3,pp. 189–201, Aug. 1998.
[21] C. Lin and K.-M. Wang, “Mobile multicast support in IP networks,” in Proceedings of the Con-ference on Computer Communications (IEEE Infocom), (Tel Aviv, Israel), Mar. 2000.
[22] C. Tan and S. Pink, “MobiCast: A multicast scheme for wireless networks,” Mobile Networks andApplications, vol. 5, no. 4, pp. 259–271, 2000.
[23] K. Kim, J. Ha, E. Hyun, and S. Kim, “New approach for mobile multicast based on ssm,” pp. 405–408, Oct. 2001.
[24] A. Dutta, O. Altintas, W. Chen, and H. Schulzrinne, “Mobility approaches for all IP wirelessnetworks,” in SCI, (Orlando, Florida), July 2002.
[25] A. Dutta, F. Vakil, J. Chen, M. Tauil, S. Baba, and H. Schulzrinne, “Application layer mobilitymanagement scheme for wireless Internet,” in 3G Wireless, (San Francisco), p. 7, May 2001.
[26] A. Dutta, B. Kim, T. Zhang, T. Technologies, S. Baba, K. Taniuchi, Y. Ohba, and H. Schulzrinne,“Experimental analysis of multi interface mobility management with sip and mip,” in IEEE Wire-less Conference, (Maui, HI), June 2005.
[27] A. Dutta, K. Manousakis, S. Das, T. Chiba, and H. Shulzrinne, “Mobilty Testbed for 3GPP2-basedMultimedia Domain Networks,” IEEE Communications Magazine, vol. 45, p. 118, July 2007.
[28] M. Greis, “3GPP TS23. 107 v5. 0.0,” Quality of Service, Concept and Architecture, April, 2001.
[29] A. Dutta, H. Schulzrinne, T. Chiba, H. Yokota, and S. Das, “Comparative Analysis of NetworkLayer and Application Layer IP Mobility Protocols for IPv6 Networks,” in Proceedings of WPMC2006, (San Diego), Sept. 2006.
[30] J. M. (Ed.) and M. K. (Ed.), “Mobility related terminology,” RFC 3753, Internet Engineering TaskForce, June 2004.
[31] J. Arkko, B. Aboba, J. Kohonen, and F. Bari, “Network discovery and selection problem,” RFC5113, Internet Engineering Task Force, Jan. 2008.
[32] J. Choi, “Goals of detecting network attachment in IPv6,” RFC 4135, Internet Engineering TaskForce, Aug. 2005.
[33] H. Fathi, K. Kobara, S. Chakraborty, H. Imai, and R. Prasad, “On the Impact of Security on theLatency in WLAN 802.11,” Nov 2005.
[34] D. Forsberg, “Protocol for carrying authentication for network access (PANA),” Internet Draftdraft-ietf-pana-pana-03, Internet Engineering Task Force, Feb. 2004. Work in progress.32
[35] M. Georgides, “Context transfer support for IP-based mobility management,” 2004.
[36] B. Aboba, L. Blunk, J. Vollbrecht, J. Carlson, and E. Levkowetz, “Extensible authentication pro-tocol (EAP),” RFC 3748, Internet Engineering Task Force, June 2004.
[37] B. Aboba and D. Simon, “PPP EAP TLS authentication protocol,” RFC 2716, Internet Engineer-ing Task Force, Oct. 1999.
[38] D. Maughan, M. Schertler, M. Schneider, and J. Turner, “Internet security association and keymanagement protocol (ISAKMP),” RFC 2408, Internet Engineering Task Force, Nov. 1998.
[39] A. Mishra, M. Shin, and W. Arbaugh, “An empirical analysis of the IEEE 802.11 MAC layerhandoff process,” ACM Computer Communication Review, vol. 33, no. 2, pp. 93–102, 2003.
[40] R. Droms, “Dynamic host configuration protocol,” RFC 2131, Internet Engineering Task Force,Mar. 1997.
[41] R. Droms, I. Ed., J. Bound, B. Volz, T. Lemon, C. Perkins, and M. Carney, “Dynamic host con-figuration protocol for IPv6 (DHCPv6),” RFC 3315, Internet Engineering Task Force, July 2003.
[42] “The Point-to-Point protocol (PPP),” RFC 1661, Internet Engineering Task Force, July 1994.
[43] S. Thomson and T. Narten, “IPv6 stateless address autoconfiguration,” RFC 2462, Internet Engi-neering Task Force, Dec. 1998.
[44] International Workshop on Petrinet and Performance Models, On Petri Net-based ModelingParadigms for the Performance Analysis of Wireless Internet Access, IEEE, 2001.
[45] R. Amadio and S. Prasad, “Modelling IP Mobility,” CONCUR’98: Concurrency Theory: 9thInternational Conference, September 1998.
[46] D. Tutsch and J. Sokol, “Petri Net based Performance Evaluation of USAIAs Bandwidth Parti-tioning for the Wireless Cell Level,” Proceedings of the 9th international Workshop on Petri Netsand Performance Models PNPM’01, Sept. 2001.
[47] C. Molina-Ramirez, D. Munoz-Rodriguez, and E. Lopez-Mellado, “Modelling and analysis oftelecommunication cellular systems usingPetri nets,” Systems, Man, and Cybernetics, IEEE In-ternational Conference on Humans, Information and Technology, vol. 3, 1994.
[48] F. Jaimes-Romero, D. Munoz-Rodriguez, C. Molina, and H. Tawfik, “Modeling resource manage-ment in cellular systems using Petri nets,” Vehicular Technology, IEEE Transactions on, vol. 46,no. 2, pp. 298–312, 1997.
[49] F. Boussinot, R. de Simone, and V. ENSMP-CMA, “The ESTEREL language,” Proceedings ofthe IEEE, vol. 79, no. 9, pp. 1293–1304, 1991.
[50] R. Cuninghame-Green, “Minimax algebra and applications,” Fuzzy Sets and Systems, vol. 41,no. 3, pp. 251–267, 1991.
[51] A. Hormozi and L. Dube, “Establishing Project Control: Schedule, Cost, and Quality.,” SAMAdvanced Management Journal, vol. 64, no. 4, 1999.
[52] A. Zimmermann, R. German, J. Freiheit, and G. Hommel, “Timenet 3.0 tool description,” 1999.
[53] C. Hirel, B. Tun, and K. Trivedi, “SPNP Version 6.0,” Computer performance evaluation: Mod-elling tools and techniques; 11th International Conference; TOOLS, pp. 354–357, 2000.
[54] W. Sanders, “UltraSAN Users Manual Version 3.0,” Center for Reliable and High-PerformanceComputing, University of Illinois, 1995. 33
[55] G. Klas and R. Lepold, “TOMSPIN-a tool for modeling with stochastic Petri nets,” Com-pEuro’92.’Computer Systems and Software Engineering’, Proceedings., pp. 618–623, 1992.
[56] W. M. Zuberek and W. Kubiak, “Timed petri nets in modeling and analysis of simple schedules formanufacturing cellss,” Elsevier Science, Computers and Mathematics with Applications, vol. 37,1999.
[57] T. Murata, “Petri nets: Properties, analysis and applications,” Proceedings of IEEE, Apr. 1989.
[58] W. Zuberek, “Hierarchical analysis of manufacturing systems using Petri nets,” vol. 4, Oct. 2000.
[59] R. Floyd, “Algorithm 97: Shortest path,” Communications of the ACM, vol. 5, no. 6, p. 345, 1962.
[60] A. Dutta, P. Agrawal, S. Das, M. Elaoud, D. Famolari, S. Madhani, A. McAuley, T. M. Li, P., andH. Schulzrinne, “Realizing mobile wireless Internet telephony and streaming multimedia testbed,”Computer Communications, vol. 27, no. 8, pp. 725–738, 2004.
[61] S. Shin, H. Schulzrinne, A. Forte, and A. Rawat, “Reducing MAC layer handoff latency in IEEE802.11 wireless LANs,” in ACM MobiWac (ACM International Workshop on Mobility Manage-ment and Wireless Access), (Philadelphia, Pennsylvania), ACM, Sept. 2004.
[62] J. Montavont, N. Montavont, and T. Noel, “Enhanced schemes for L2 handover in IEEE 802.11networks and their evaluations,” in 16th International Symposium on Personal Indoor and MobileRadio Communications, (Berlin, Germany), Sept. 2005.
[63] H. Velayos and G. Karlsson, “Techniques to reduce IEEE 802.11b handoff time,” in WirelessNetworking Symposium, (Paris, France), June 2004.
[64] V. Brik, A. Mishra, and S. Banerjee, “Eliminating handoff latencies in 802.11 WLANs usingmultiple radios: Applications, experience, and evaluation,” in ACM SIGCOMM IMC, (Berkeley,CA), Oct. 2005.
[65] A. Forte and H. Schulzrinne, “Cooperation Between Stations in Wireless Networks,” in NetworkProtocols, 2007. ICNP 2007. IEEE International Conference on, pp. 31–40, 2007.
[66] A. Dutta, S. Madhani, T. Zhang, Y. Ohba, and K. Taniuchi, “Network discovery mechanisms forfast-handoff,” in Broadnets, (San Jose), IEEE, 2006.
[67] T. Zhang, S. Madhani, E. AshutoshDutta, Y. Ohba, K. Taniuchi, and S. Mohanty, “Implementationand evaluation of autonomous collaborative discovery of neighboring networks,” in InformationTechnology: Research and Education, 2005. ITRE 2005. 3rd International Conference on, pp. 12–17, 2005.
[68] A. Dutta, S. Das, D. Famolari, Y. Ohba, K. Taniuchi, V. Fajardo, T. Kodama, and H. Schulzrinne,“Secured seamless convergence across heterogeneous access networks,” in World Telecommuni-cation Congress, (Budapest), IEEE, May 2006.
[69] M. S. Bargh, B. Hulsebosch, H. Eertink, A. Prasad, H. Wang, and P. Schoo, “Fast authenticationmethods for handovers between IEEE 802.11 wireless LANs,” in The Second ACM InternationalWorkshop on Wireless Mobile Applications and Services on WLAN Hotspots, (Philadelphia, PA,),Oct. 2004.
[70] R. Housely and B. Aboba, “Guidance for authentication, authorization, and accounting (aaa) keymanagement,” RFC 4962, Internet Engineering Task Force, July 2007.
[71] A. Dutta, Y. Ohba, V. Fajardo, K. Taniuchi, and H. Schulzrinne, “A Framework of Media-Independent Pre-Authentication (MPA),” Internet Draft draft-irtf-mobopts-mpa-framework-02,Internet Engineering Task Force, Feb. 2008. Work in progress.34
[72] R. Lopez, A. Dutta, Y. Ohba, and H. Schulzrinne, “Network-layer assisted mechanism to opti-mize authentication delay during handoff in 802.11 networks,” in ACM Mobiquitous, (Philadel-phia,PA), June 2007.
[73] P. Calhoun, J. Loughney, E. Guttman, G. Zorn, and J. Arkko, “Diameter base protocol,” RFC3588, Internet Engineering Task Force, Sept. 2003.
[74] A. Miu and P. Bahl, “Dynamic host configuration for managing mobility between public andprivate networks,” in The 3rd Usenix Internet Technical Symposium, (San Francisco, CA), Mar.2001.
[75] P. Rodriguez, R. Chakravorty, J. Chesterfield, I. Pratt, and S. Banerjee, “MAR: a commuter routerinfrastructure for the mobile internet,” in Proceedings of the 2nd international ACM conferenceon Mobile systems, applications, and services, (Boston, MA), pp. 217–230, June 2004.
[76] S. Kent and R. Atkinson, “IP encapsulating security payload (ESP),” RFC 2406, Internet Engi-neering Task Force, Nov. 1998.
[77] A. Dutta, T. Zhang, S. Madhani, K. Taniuchi, K. Fujimoto, H. Schulzrinne, Y. Ohba, and Y. Kat-sube, “Secure universal mobility for wireless internet,” in The Second ACM International Work-shop on Wireless Mobile Applications and Services on WLAN Hotspots, (Philadelphia, PA,), Oct.2004.
[78] J.-O. Vatn and G. C. Maguire, “The effect of using co-located care-of addresses on macro han-dover latency,” in 14th Nordic Tele-traffic Seminar, (Technical University of Denmark, Lyngby,Denmark), Aug. 1998.
[79] N. Moore, “Optimistic duplicate location detection DAD for IPv6,” RFC 4429, Internet Engineer-ing Task Force, Apr. 2006.
[80] Y. Han et al., “Advance duplicate address detection,” internet draft, Internet Engineering TaskForce, July 2003. Work in progress.
[81] A. Forte, S. Shin, and H. Schulzrinne, “Passive duplicate address detection for dynamic host con-figuration protocol (DHCP),” Tech. Rep. cucs-011-06, Columbia University, Computer ScienceDepartment, Mar. 2006.
[82] S. Park, P. Kim, and B.Volz, “Rapid commit option for the dynamic host configuration protocolversion 4,” RFC 4039, Internet Engineering Task Force, 2005.
[83] T. Narten, E. Nordmark, and W. Simpson, “Neighbor discovery for IP version 6 (IPv6),” RFC2461, Internet Engineering Task Force, Dec. 1998.
[84] A. J. Tuominen and H. L. Petander, “MIPL mobile IPv6 for linux in HUT campus network medi-apoli,” in Proceedings of Ottawa Linux Symposium, 2001, (Ottawa, Canda), June 2001.
[85] N. Nakajima, A. Dutta, S. Das, and H. Schulzrinne, “Handoff delay analysis and measurementfor SIP based mobility in IPv6,” in ICC 2003 - Personal Communication Systems and WirelessLANs, May 2003.
[86] IEEE, GPS assisted Fast-handoff Mechanism for Real-Time Communication, (Sarnoff Sympo-sium, Princeton, NJ), IEEE, 2006.
[87] H. Schulzrinne and R. Hancock, “GIMPS: general internet messaging protocol for signaling,” In-ternet Draft draft-ietf-nsis-ntlp-15, Internet Engineering Task Force, Feb. 2008. Work in progress.
[88] A. Dutta, T. Zhang, Y. Ohba, K. Taniuchi, and H. Schulzrinne, “MPA assisted proactive handoffscheme,” in ACM Mobiquitous, SIGMOBILE, ACM, 2005.35
[89] C. Wu, A. Cheng, S. Lee, J. Ho, and D. Lee, “Bi-directional route optimization in mobile IP overwireless LAN,” in Vehicular Technology Conference, 2002. Proceedings. VTC 2002-Fall. 2002IEEE 56th, vol. 2, 2002.
[90] E. Wedlund and H. Schulzrinne, “Mobility support using SIP,” in 2nd ACM/IEEE InternationalConference on Wireless and Mobile Multimedia (WoWMoM), (Seattle, Washington), Aug. 1999.
[91] A. Dutta, P. Agrawal, J.-C. Chen, S. Das, D. Famolari, Y. Ohba, S. Baba, A. McAuley,and H. Schulzrinne, “Realizing steaming wireless internet telephony and streaming multimediatestbed,” Elsevier, Computer and Communication, Oct. 2003.
[92] A. Dutta, S. Das, P. Li, A. McAuley, Y. Ohba, S. Baba, and H. Schulzrinne, “Secured mobile mul-timedia communication for wireless internet,” in International Conference on Network Sensingand Control, (Taipei, Taiwan), IEEE, Mar. 2004.
[93] A. Dutta, J. Burns, R. Jain, D. Wong, K. Young, and H. Schulzrinne, “Implementation and Per-formance Evaluation of Application layer MIP-LR,” in Wireless Networks, Communications andMobile Computing, 2005 International Conference on, vol. 2, (Maui, HI), June 2005.
[94] T. Chiba, A. Dutta, and H. Schulzrinne, “Trombone Routing Mitigation Techniques forIMS/MMD Networks,” in Proceedings of IEEE WCNC, (Hong Kong), Mar. 2007.
[95] E. Fogelstroem, A. Jonsson, and C. Perkins, “Mobile IPv4 regional registration,” RFC 4857,Internet Engineering Task Force, 2007.
[96] H. Soliman, C. Castelluccia, K. el Malki, and L. Bellier, “Hierarchical mobile IPv6 mobilitymanagement HMIPv6,” RFC 4140, Internet Engineering Task Force, Aug. 2006.
[97] A. Dutta, S. Madhani, W. Chen, O. Altintas, and H. Schulzrinne, “Fast-handoff schemes forapplication layer mobility management,,” in IEEE PIMRC, Sept. 2004.
[98] E. R. Koodli, “Fast handovers for mobile IPv6,” RFC 4068, Internet Engineering Task Force, July2005.
[99] K. E. Malki, “Low-latency handoffs in mobile IPv4,” RFC 4881, Internet Engineering Task Force,June 2007.
[100] C. E. Perkins and K.-Y. Wang, “Optimized smooth handoffs in mobile IP,” in Fourth IEEE Sym-posium on Computers and Communications (ISCC ’99), (Red Sea, Egypt), pp. 340–346, July1999.
[101] P. Calhoun, T. Hiller, J. Kempf, P. McCann, C. Pairla, A. Singh, and S. Thalanany, “Foreign agentassisted hand-off,” Internet Draft, Internet Engineering Task Force, Nov. 2000. Work in progress.
[102] F. Vakil, D. Famolari, S. Baba, and D. Famolari, “Virtual soft hand-off in IP-Centric wirelessCDMA networks,” in 3G Wireless 2001, (San Francisco), Delson, May 2001.
[103] A. Dutta, H. Schulzrinne, S. Madhani, O. Altintas, and W. Chen, “Optimized fast-handoffschemes for application layer mobility management,” Mobile Computing and CommunicationsReview (MC2R), vol. Volume 6, Nov. 2002.
[104] H. Schulzrinne, “SIP registration,” internet draft, Internet Engineering Task Force, Apr. 2001.Work in progress.
[105] H. Schulzrinne, S. Casner, R. Frederick, and V. Jacobson, “RTP: a transport protocol for Real-Time applications,” RFC 3550, Internet Engineering Task Force, July 2003.
[106] H. Schulzrinne, “RTPTrans http://www. cs. columbia. edu,” IRT/software/rtptools.36
[107] G. Herrin, “Linux IP Networking: A Guide to the Implementation and Modification of the LinuxProtocol Stack,” Connections, vol. 3, p. 2, 2000.
[108] J. Lennox and H. Schulzrinne, “Transporting user control information in SIP REGISTER pay-loads,” internet draft, Internet Engineering Task Force, Mar. 1999. Work in progress.
[109] M. A. Sasse, V. J. Hardman, I. Kouvelas, C. E. Perkins, O. Hodson, A. I. Watson, M. Handley,and J. Crowcroft, “RAT (robust-audio tool),” 1995.
[110] A. Misra, S. Das, A. Dutta, A. McAuley, and S. Das, “IDMP based fast-handoff and paging in IPbased 4G mobile networks,” IEEE Communications Magazine, vol. 40, pp. 138–145, Mar. 2002.
[111] N. Moore, “Edge handovers for mobile IPv6,” Internet Draft draft-moore-mobopts-edge-handovers-00, Internet Engineering Task Force, Feb. 2004. Work in progress.
[112] G. Krishnamurthi, R. Chalmers, and C. Perkins, “Buffer management for smooth HandOvers inmobile IPv6,” internet draft, Internet Engineering Task Force, Mar. 2001. Work in progress.
[113] J. Rosenberg and H. Schulzrinne, “An RTP payload format for generic forward error correction,”RFC 2733, Internet Engineering Task Force, Dec. 1999.
[114] J. Ott, S. Wenger, N. Sato, C. Burmeister, and J. Rey, “Extended rtp profile for real-time transportcontrol protocol (rtcp)-based feedback (rtp/avpf),” RFC 4585, Internet Engineering Task Force,July 2006.
[115] C. Perkins and O. Hodson, “Options for repair of streaming media,” RFC 2354, Internet Engi-neering Task Force, June 1998.
[116] A. Dutta, E. van den Berg, D. Famolari, V. Fajardo, Y. Ohba, K. Taniuchi, and H. Schulzrinne,“Dynamic buffering scheme for mobile handoff,” in PIMRC, (Helsinki), IEEE, 2006.
[117] K. Wong, H.-Y. Wei, A. Dutta, K. Young, and H. Schulzrinne, “IP micro-mobility managementusing host-based routing,” in Wireless IP and building the Mobile Internet (S. Dixit and R. Prasad,eds.), Artech House, 2002.
[118] A. Valko, “Cellular IP: a new approach to Internet host mobility,” ACM SIGCOMM ComputerCommunication Review, vol. 29, no. 1, pp. 50–65, 1999.
[119] A. Dutta, D. Wong, J. Burns, R. Jain, K. Young, A. McAuley, and H. Schulzrinne, “Realization ofintegrated mobility management for ad-hoc networks,” in IEEE Milcom, (Annaheim, California),Oct. 2002.
[120] D. Wong, A. Dutta, J. Burns, K. Young, and H. Schulzrinne, “A multilayered mobility manage-ment scheme for auto-configured wireless IP networks,” IEEE Wireless Communication Maga-zine, vol. 10, Oct. 2003.
[121] J. K. Ed., “Goals for network based licalized mobility management,” RFC 4830, Internet Engi-neering Task Force, Apr. 2007.
[122] Y. Gwon, G. Fu, and R. Jain, “Fast handoffs in wireless LAN networks using mobile initiated tun-neling handoff protocol for ipv4 MITHv4,” in IEEE Wireless Communications and Networking,pp. 1248–1253, Mar. 2003.
[123] A. Dutta, S. Das, D. Famolari, Y. Ohba, and H. Schulzrinne, “Seamless Proactive Handoveracross Heterogeneous Access Networks,” Wireless Personal Communication, vol. 43, pp. 837–855, August 2007.
37
[124] A. Zimmermann, J. Freiheit, R. German, and G. Hommel, “Petri net modeling and performabil-ity evaluation with timeNET 3.0.,” Lecture Notes in Computer Science: Computer PerformanceEvaluation: Modeling Techniques and Tools, vol. 1786, pp. 188–202, 2000.
38
(List of my Publication - categorized)
Mobility Fast-Handoff - Techniques, Systems prototype and Experiments
1. A. Dutta, S. Madhani, H. Schulzrinne, O. Altintas, W. Chen, “Optimized Fast-handoff Schemesfor Application Layer Mobility Management,” ACM MC2R, Vol 7, Issue 1, January 2003
2. N. Nakajima, A. Dutta, S. Das, H. Schulzrinne, “Handoff Delay Analysis for SIP Mobility inIPv6 Testbed,” IEEE ICC May, 2003, Alaska
3. P-yu Hsieh, A. Dutta, H. Schulzrinne, “Application Layer Mobility Proxy for Real-time commu-nication,” 3G Wireless, May 2003, San Francisco
4. A. Misra, S. Das, A. Dutta, A. McAuley, S. Das, “IDMP based Fast-handoff and Paging in IPbased 4G Mobile Networks,” IEEE Communication Magazine, March 2002
5. K. Daniel Wong, A. Dutta, K. Young, H. Schulzrinne, “Managing Simultaneous Mobility of IPHosts,” IEEE MILCOM 2003, Boston
6. A. Dutta, J. Chennikara, W. Chen, O. Altintas, H. Schulzrinne, “Multicasting streaming media tomobile users,” IEEE Communication Magazine, October 2003 Issue
7. D. Wong, A. Dutta, J. Burns, R. Jain, K. Young, H. Schulzrinne, “A multilayered mobility man-agement scheme for auto-configured wireless networks,” IEEE Wireless Communication, October2003 Issue
8. A. Dutta, S. Madhani, W. Chen, O. Altintas, H. Schulzrinne, “GPS-assisted Fast-handoff forReal-time communication,” IEEE Sarnoff Symposium 2006, Princeton, NJ
9. A. Dutta, S. Das, P. Li, A. McAuley, Y. Ohba, S. Baba, H. Schulzrinne, “Secured Mobile Multi-media Communication for Wireless Internet,” IEEE ICNSC 2004, Taipei, Taiwan
10. A. Dutta, H. Schulzrinne, “MarconiNet: Overlay Mobile Content Distribution Network,” IEEECommunication Magazine, February 2004 Issue
11. A. Dutta, S. Madhani, W. Chen, O. Altintas, H. Schulzrinne, “Fast-handoff Schemes for Applica-tion Layer Mobility Management,” IEEE PIMRC 2004, Barcelona
12. A. Dutta, T. Zhang, S. Madhani, K. Taniuchi, K. Fujimoto, Y. Katsube, Y. Ohba, H. Schulzrinne,“Secure Universal Mobility for Wireless Internet,” ACM WMASH 2004, Philadelphia
13. D. Wong, A. Dutta, Simultaneous Mobility in MIPv6, IEEE, EIT 2005, Nebraska. (Best PaperAward)
14. A. Dutta, T. Zhang, S. Madhani, K. Taniuchi, K. Fujimoto, Y. Katsube, Y. Ohba, H. SchulzrinneSecure Universal Mobility for Wireless Internet, (Extended version) ACM MC2R, July 2005
15. A. Dutta, T. Zhang, K. Taniuchi, Y.Ohba, H. Schulzrinne, “MPA assisted Optimized ProactiveHandoff Scheme”, ACM Mobiquitous 2005, San Diego, CA
16. A. Dutta, S. Das, D. Famolari, Y. Ohba, K. Taniuchi, T. Kodama, H. Schulzrinne, “SeamlessHandoff across Heterogeneous Networks - An 802.21 Centric Approach,” IEEE WPMC 2005,Aalborg, Denmark
17. K.D Wong, A. Dutta, H. Schulzrinne, K. Young Simultaneous Mobility: Analysis Framework,Theorems and Solutions, Wiley Wireless Communications and Mobile Computing Journal
39
18. A. Dutta, S. Das, D. Famolari, Y. Ohba, K. Taniuchi, V. Fajardo, T. Kodama, H. Schulzrinne,“Secured Seamless Convergence across Heterogeneous Access Networks,” World Telecommuni-cation Congress 2006, Budapest, Hungary
19. A. Dutta, H. Schulzrinne, K.D Wong, “Supporting Continuous Services to Roaming Clients, TheHandbook of Mobile Middleware, CRC Press, Edited by Bellavista and Corradi.
20. A. Dutta, S. Madhani, T. Zhang, Y. Ohba, K. Taniuchi, H.Schulzrinne, “Network DiscoveryMechanism for Fast-handoff,” IEEE Broadnets 2006, San Jose.
21. A. Dutta, E. Van den Berg, D. Famolari, V. Fajardo, Y.Ohba, K. Taniuchi, T. Kodama, H.Schulzrinne,“Dynamic Buffering Control Scheme for Mobile Handoff,” IEEE PIMRC 2006, Helsinki
22. T.Chiba, H.Yokota, A.Idoue, A. Dutta, K. Manousakis, S. Das, H. Schulzrinne, “Trombone Rout-ing Mitigation Techniques for IMS/MMD Networks,” IEEE WCNC 2007, Hong Kong
23. Rafa M. Lopez,A. Dutta, Y. Ohba, H. Schulzrinne, A.F Gomez Skarmeta, “Network-Layer As-sisted Mechanism to Optimize Authentication Delay Handoff in 802.11 Networks,” ACM Mobiq-uitous, Aug. 2007, Philadelphia
24. A. Dutta, S. Das, D. Famolari, Y. Ohba, K. Taniuchi, V. Fajardo, R. M. Lopez, T. Kodama,H. Schulzrinne, “Seamless proactive handover across heterogeneous access networks,” WirelessPersonal Communication, Springer Journal
25. A. Dutta, S. Chakravarty, K. Taniuchi, V. Fajardo, Y. Ohba, D. Famolari, H. Schulzrinne, “AnExperimental Study of Location Assisted Proactive Handover,” IEEE Globecom 2007, InternetProtocol Symposium, Washington D.CMobility Fast-Handoff - Modeling
26. A. Dutta, B. Lyles, H. Schulzrinne, T. Chiba, H. Yokota, A. Idoue, “Generalized Modeling Frame-work for Handoff Analysis,” IEEE PIMRC, September 2007, AthensPerformance - Mobility
27. K. D Wong, H-Yu Wei, A. Dutta, K. Young, H. Schulzrinne, “Performance of IP Micro-MobilityManagement Scheme using Host Based Routing”, IEEE WPMC 2001, Aalborg
28. F. Anjum, M. Elaoud, D. Famolari, A. Ghose, R. Vaidyanathan, A. Dutta, P. Agrawal, “VoicePerformance in WLAN Networks - An Experimental Study,” IEEE Globecom 2003, San Francisco
29. A. Dutta, J. Burns, R. Jain, K. D. Wong, K. Young, H. Schulzrinne, “ Implementation and Perfor-mance Evaluation of Application layer MIP-LR,” IEEE Wirelesscom 2005, Maui, HI
30. A. Dutta, B. Kim, T. Zhang, S. Baba, K. Taniuchi, Y.Ohba, H.Schulzrinne, “Experimental Anal-ysis of Multi Interface Mobility Management with SIP and MIP,” IEEE Wirelesscom 2005, Maui,HI
31. A. Dutta, S. Chakrabarty, K. Taniuchi, V. Fajardo, Y.Ohba, D. Famolari, H. Schulzrinne, ”AnExperimental Study of Location Assisted Proactive Handover, IEEE Globecom 2007, WashingtonDCMobility and Streaming Architecture
32. A. Dutta, H. Schulzrinne, Y. Yemini, ”MarconiNet: An Architecture for Internet Radio and TV.9th International Workshop on Network Support for Digital Audio Video Systems (NOSSDAV99), New Jersey, 23-25th June
40
33. A. Dutta, H. Schulzrinne, “A Streaming Architecture for Next Generation Internet,” IEEE ICC2001, June 11-14, 2001, Helsinki, Finland
34. A. Dutta, F. Vakil, J.C Chen, M. Tauil, S. Baba, H. Schulzrinne, ”Application Layer MobilityManagement Scheme for Wireless Internet,” in 3G Wireless May 2001,(San Francisco)
35. K. Chakrabarty, A. Misra, S. Das, A. McAuley, A. Dutta, S. Das, ”Implementation and Perfor-mance Evaluation of TeleMIP,” IEEE ICC, May 2001, Helsinki
36. A. Misra, S. Das, A. Dutta, A. McAuley, S. Das, ”IDMP-based Fast handoffs and paging inIP-based cellular Networks,” in 3GWireless 2001, May 2001, San Francisco
37. S. Das, A. Misra, A. McAuley, A. Dutta, S. Das, ”A generalized mobility solution using a dynamictunneling,” in ICCCD2000, Dec. 2000, Kharagpur, India
38. A. Misra, S. Das, A. McAuley, A. Dutta, S. Das, “Integrating QoS Support in TeleMIP’s MobilityArchitecture,” in ICPWC, (Hyderabad, India), pp. 8, Dec. 2000
39. A. Dutta, O. Altintas, H. Schulzrinne, W. Chen, “Multimedia SIP sessions in a Mobile Heteroge-neous Access Environment,” 3G Wireless 2002, San Francisco
40. J. Chennikara, W. Chen, A. Dutta, O. Altintas, “Application Layer Multicast for Mobile Users inDiverse Networks,” IEEE Globecom 2002, Taiwan
41. S. Das, A. Dutta, A. McAuley, A. Misra, S. Das, “IDMP: An Intra-Domain Mobility ManagementProtocol for Next Generation,” IEEE Wireless Magazine, October 2002 - SAIC Best Paper
42. T. Chiba, H. Yokota, A. Idoue, A. Dutta, S. Das, Fuchun J. Lin, H.Schulzrinne, “Mobility Man-agement Schemes for Heterogeneity Support in Next Generation Wireless Networks,” NGI 2007,May 2007, Norway
43. A. Dutta, H. Schulzrinne, W. Chen, O. Altintas, ”Mobility support for wireless streaming multi-media in MarconiNet,” in IEEE Broadband Wireless Summit, Interop 2001, (Las Vegas), pp. 7,May 2001
44. A. Dutta, H. Schulzrinne, S. Das, A. McAuley, W. Chen, O. Altintas, “ MarconiNet support-ing Streaming Media over Localized Wireless Multicast,” ACM M-Commerce 2002 Workshop,September, 2002, Atlanta
45. A. Dutta, R. Jain, K. D. Wong, J. Burns, K. Young, Henning Schulzrinne, “Multilayerd MobilityManagement for Survivable Network,” IEEE MILCOM Proceedings, October 2001, Boston
46. Y. Ohba, S. Das, A. Dutta, “Kerberized Handover Keying: A Media-Independent Handover KeyManagement Architecture,” ACM Mobiarch 2007, Kyoto, Japan
47. A. Dutta, F. Joe Lin, D. Chee, S. Das, B. Lyles, T. Chiba, H. Yokota, H. Schulzrinne ”ArchitectureAnalysis and Experimental IPv6 tstbed for Advances IMS,” IMSAA 2007, Bangalore, IndiaTestbed - Mobility
48. A. Dutta, J.C Chen, S. Das, S. Madhani, A. McAuley, S. Baba, N. Nakajima, Y. Ohba, H.Schulzrinne, ”Implementing a Testbed for Mobile Multimedia,” IEEE Globecom 2001, San An-tonio
49. A. Dutta, D. Wong, J. Burns, R. Jain, H. Schulzrinne, A. McAuley, “Realization of IntegratedMobility Management for Ad-Hoc Networks,” IEEE MILCOM 2002 , Anaheim, CA
41
50. A. Dutta, P. Agrawal, S. Das, M. Elaoud, D. Famolari, H. Schulzrinne et al, “Realizing MobileWireless Internet Telephony and Streaming Multimedia Testbed,” Elsevier Journal for Computerand Communication, April 2001
51. A. Dutta, K. Manousakis, S. Das, F.J Lin, T. Chiba, H. Yokota, A. Idoue, H.Schulzrinne, “Mobil-ity Testbed for 3GPP2-based MMD Networks,” IEEE Communication Magazine, July 2007Survey Paper - Mobility
52. A. Dutta, O. Altintas, W. Chen, H. Schulzrinne, “Mobility Approaches for All IP Wireless Net-works,” SCI 2002, Orlando, Florida.
53. T. Chiba, H. Yokota, A. Idoue, A. Dutta, S. Das, Fuchun J. Lin, “Gap Analysis and DeploymentArchitectures for 3GPP2 MMD Networks,” IEEE VT Magazine, March 2007.
54. A. Dutta, S. Das, T. Chiba, H. Yokota, A. Idoue, H. Schulzrinne, “Comparative Analysis ofNetwork Layer and Application Layer IP Mobility Protocols for IPv6 Networks,” WPMC 2006,San Diego, CAPotpourri
55. A. Dutta, Y. Yemini, ”Power Management of LEOs under bursty broadband traffic. AIAA’s 17thInternational Conference on Satellite Systems and Communication. AIAA, 1998 February,” Yoko-hama, Japan, March 1998
56. S. Khurana, A. Dutta, P. Gurung, H. Schulzrinne, “XML based Wide Area Communication withNetworked Appliances,” IEEE Sarnoff 2004, Princeton, NJ
57. J. Chennikara, A. Dutta, A. McAuley, D. Wong, M. Elaoud, A. Cheng, M. Yajnik, I. Sebuktekin,K. Young, H. Schulzrinne, “ Integrated Networking Technologies for Survivable Network,” IEEEWCNC, March 2005, New Orleans
58. T. Zhang, S. Madhani, A. Dutta, E. Van den Berg, Y.Ohba, K. Taniuchi, S. Mohanty, “ Implemen-tation and Evaluation of Autonomous Collaborative Discovery of Neighboring Networks,” IEEEITRE 2005
59. A. Dutta, H. Cheng, S. Madahani, K.D. Wong, J. Chennikara, K. Young, H. Schulzrinne, A. Patel,“Flexible Call Control Framework for Supporting Multi-party Service,” IEEE MILCOM 2005
60. A. Dutta, J. Alberi, A. Cheng, B. Horgan , A. McAuley, D. Chee, B. Lyles, “IPv6 TransitionTechniques for Legacy Application,” IEEE MILCOM 2006
42