Understanding and Enhancing the Understanding and Enhancing the Scalability of IMSScalability of IMS--based based
Infrastructures and ServicesInfrastructures and Services
Luca Foschini
Oct. 20th 2009, P2MNet, Zurich
Research FellowDEIS, University of Bologna, ITALY
Zurich — 20.10.2009 P2MNet 2/54
AgendaAgenda
Service delivery in 4G converged Internet
IP Multimedia Subsystem – IMS – scalabilityissues
IHMAS middleware for scalability of IMS-based infrastructure and services
IHMAS presence service scalability use case– Infrastructure scalability
– Intra-/Inter-domain service scalability
– Implementation details and experimental evaluation
Concluding remarks and open research directions
Zurich — 20.10.2009 P2MNet 3/54
Ever-increasing demand and diffusion of mobile multimedia services during the last two decades, driven by:
– New powerful devices and wireless technologies/infrastructures
– New (mobile) services
4G converged world4G converged world
Voice (VoIP)
Audio/Video Conference
Video on Demand (VoD)
Chat and messagging
And many more…
?...
- Push To Talk (PTT)- PTT over Cellular (PoC)- IPTV - Video sharing- ...
Zurich — 20.10.2009 P2MNet 4/54
4G converged 4G converged service delivery scenarioservice delivery scenario
Mobile multimedia services offered
by network operators(e.g., VoIP, IPTV, …)
Mobile multimedia services offered by third party (Internet)
service providers(e.g., weather forecast, news, …)
Operators’ core IP networks providing basic services:QoS-enabled data transport, mobility, AAA, …
Service delivery platform: an all-IP overlay to
facilitate service access and integration (e.g. IMS)
IEEE 802.11(WiFi)
Bluetooth
Cellular 3G and WiMAX
Highly differentiated (wireless) access networks
Zurich — 20.10.2009 P2MNet 5/54
IP-basednetwork
Hotspot Wi-Fi Hotspot Wi-Fi
3G cellular
4G service & network management:4G service & network management:a proxya proxy--based approachbased approach
IMSIP Multimedia Subsystembased on SIP(Session Initiation Protocol)
New protocols and active proxy-based infrastructures for session management
Zurich — 20.10.2009 P2MNet 6/54
New service scenarios in 4G:New service scenarios in 4G:handoff management of mobile multimediahandoff management of mobile multimedia
WLAN
Bluetooth GPRS
IP-basednetwork
IMS-based components
Zurich — 20.10.2009 P2MNet 7/54
PSP
WatchersIMS-based
Presence ServicePresentities
Presence service (PS) permits users and hw/sw components, called presentities (Pi ), to convey their ability and willingness to communicate with subscribed watchers (Wj )
PUBLISH IMSP
New support services in 4G:New support services in 4G:presence servicepresence service
P1
P2
PNWN
W2
W1
SUBSCRIBE(to P2@domainP)
SUBSCRIBE
SUBSCRIBE
(to P2@domainP)
(to P2@domainP)
IMSW NOTIFY
NOTIFY
domainP domainWNOTIFY
Zurich — 20.10.2009 P2MNet 8/54
Need for a better understanding of IMS scalability shortcomings and load-balancing support both at infrastructure and service levels
New services VoIP+PS(call-status notification)
Higher signaling traffic (message dimension + frequency)Richer services, such as VoIP+PS(message multiplying effect)Many traversed signaling entities (proxies-based architecture…)Plus, specific SIP protocol issues (message verbosity and ACKs)
Scalability issues at a glanceScalability issues at a glance
High mobility & context changes
Zurich — 20.10.2009 P2MNet 9/54
Some background:Some background:SIP SIP –– SessionSession InitiationInitiation ProtocolProtocol
SIP defines a signaling framework and related protocols and messages to setup any kind of session(work at the Open Systems Interconnection – OSI –session layer)
– SIP is very open and general purpose ☺– SIP includes several core facilities for mobility management,
session initiation, termination, and transfer, …– SIP does not include some basic services
(e.g., AAA, resource booking, …)
SIP is not a data/media transmission protocolOther specific protocols for that: Real-time Transport Protocol (RTP), RTP Control Protocol (RTCP), Real Time Streaming (RTSP),…
SIP usage examples– Setting up and tearing down VoIP voice calls– Instance messaging and presence service: SIP for Instant
Messaging and Presence Leveraging Extensions – SIMPLE– Session transfer and call re-direction
Zurich — 20.10.2009 P2MNet 10/54
SIP in a nutshellSIP in a nutshell
SIP core signaling– HTTP-like text-based protocol and email-like SIP identifiers (addresses)
– Client/server protocol (request/response protocol)– Standardized session control messages
INVITE, REGISTER, OK, ACK, BYE, …
SIP proxy-based framework and main entities– User agents: end points, can act as both user agent client and as
user agent serverUser Agent Client: create new SIP requestsUser Agent Server: generate responses to SIP requests
– Dialog: peer to peer relationship between two user agents, established by specific methods
– Proxy servers: application level routers– Redirect servers: redirect clients to alternate servers– Registrars: keep tracks of users
Zurich — 20.10.2009 P2MNet 11/54
SIP SIP VoIPVoIP call initiation example:call initiation example:INVITE dialogINVITE dialog
CALLER PROXY A CALLEEPROXY B
(1) INVITE
(4) ACK
(6) 200 OK
Media session (e.g. RTP-based, NOT SIP)
(2) RINGING
(3) 200 OK
(5) BYE
INVITE sip:[email protected] SIP/2.0
Via: SIP/2.0/UDP pc33.atlanta.com;branch=z9hG4bK776asdhds
Max-Forwards: 70
To: Bob <sip:[email protected]>
From: Alice <sip:[email protected]>;tag=1928301774
Call-ID: [email protected]
CSeq: 314159 INVITE
Contact: <sip:[email protected]>
Content-Type: application/sdp
Content-Length: 142
… SDP description …
Start line:• request line (in requests)• status line (in responses)
Header: with a number of header fields
Message body (optional): for example, an SDP description to negotiate audio/video codecs/formats
SIP is very verbose and
acknowledged; in addition,
SIP parsing is CPU-intensive
Zurich — 20.10.2009 P2MNet 12/54
DNS1
Some background:Some background:IMS IMS –– IP Multimedia SubsystemIP Multimedia Subsystem
Visited network 1
P-CSCF
Home Network 1 Home Network 2
Visited network 2
P-CSCF
AS AS
S-CSCF S-CSCF
I-CSCF I-CSCFHSS HSS
MobileNode(MN)
Corresp. Node(CN)
Proxy-based architectureworking at OSI session layer
Common set of functions to easemultimedia service deployment
in highly heterogeneouswireless computing environments
SIP (+ Diameter)
DATA FLOW
Signaling Functions
Application Servers (ASs)Signaling Functions/Entities for
IMS Signaling Extension/Integration
Authentication Functions
DNS2
Zurich — 20.10.2009 P2MNet 13/54
IMS functional entities:IMS functional entities:DNS and HSSDNS and HSS
Domain Name System (DNS):Standard Internet naming serviceEmployed by IMS to resolve the IP addresses of CSCFs and ASs
can be used for load balancing ☺(but… only with limited DNS-query frequency)
Home Subscriber Server (HSS):SIP requests forwarding in the appropriate direction(terminals or IMS network)
Storage of all user-related subscription data, such as authentication data and profiles for clients (by using standard Data Base Management System – DBMS)
A network may contain one or several– Subscriber Location Function (SLF) to map users to specific HSS
Zurich — 20.10.2009 P2MNet 14/54
IMS functional entities:IMS functional entities:ProxyProxy--CSCFCSCF
Proxy-Call Session Control Function (P-CSCF):First contact point in the IMS network in either visited domain or home domain
Outbound / In-bound SIP proxy(all requests from/to IMS terminals go through it)
Main P-CSCF functionsSIP requests forwarding in the appropriate direction(terminals or IMS network)
Several other functions:– Security
– Generation of charging information
– Compression and decompression of messages
Zurich — 20.10.2009 P2MNet 15/54
IMS functional entities:IMS functional entities:InterrogatingInterrogating--CSCFCSCF
Interrogating-Call Session Control Function (I-CSCF):SIP proxy at the edge of the administrative home domain
– There may be several in the same network for scalability reasons– Listed in the domain name server (DNS-based scalability)
SIP redirect stateless server
Main I-CSCF functionsInteraction with HSS to determine the S-CSCF associated with the client (Diameter protocol)
Redirection and routing of incoming SIP requests to S-CSCF
can be used to dynamically select less-loaded S-CSCFs (e.g. through DNS) ☺
Zurich — 20.10.2009 P2MNet 16/54
IMS functional entities:IMS functional entities:ServingServing--CSCFCSCF
Serving-Call Session Control Function (S-CSCF):Always located in home domainSIP proxy + SIP registrar with possibility of performing session control
Main S-CSCF functionsBinding between IP address (terminal location) and user SIP addressInteraction with application servers for value added service purposeTranslation services (Telephone number / Sip URIs)Message routing (by using so-called IMS filtering criteria)
can be used to statically divide incoming load according to user identity/profile ☺
Zurich — 20.10.2009 P2MNet 17/54
IMS functional entities:IMS functional entities:ASAS
Application Server (AS):Host services and execute services
Communicates using SIP: very costly!!– Each interposed AS generates 2 msgs (processed+ACK)
– Complex coordination for stateful and distributed ASs
Several AS types with different functionsSIP AS: signaling specific architecture (services can work only in SIP environment)
Other types: Open Service Architecture – Service Capability Server (OSA/SCS), IP Multimedia Service Switching Function (IM-SSF), …
Zurich — 20.10.2009 P2MNet 18/54
IMS RevisitedIMS Revisited
Visited network 1
P-CSCF
Home Network 1 Home Network 2
Visited network 2
P-CSCF
AS AS
S-CSCF S-CSCF
I-CSCF I-CSCFHSS HSS
MobileNode(MN)
Corresp. Node(CN)
DATA FLOW
DNS1 DNS2
1.
4.
5.
6.
7.
8.2.MN “in call”
3.
I-CSCF requiredon registration
load-balancing
Many components high overhead
Adequate infrastructure load-balancing
support, BUT no monitoring support
AS load-balancing support is insufficient
Zurich — 20.10.2009 P2MNet 19/54
Single host (local) optimizations w/out (or with minimal) coordination:– Selective message dropping– SIP message compression and
incremental parsing techniques– Stateful vs Stateless SIP proxies
Intra-domain (distributed) load-balancing:– Infrastructure-level monitoring and
dynamic load-balancing operations– Service-level AS coordination protocols
(also ad-hoc and NON-IMS-compliant optimized protocols!!)
Inter-domain protocol optimizations:– Limit traffic among different domains– Service-level message processing at IMS
domain borders (BUT, IMS compliant)
IMS scalability:IMS scalability:(partial) solutions(partial) solutions
Widely diffused and standardized
Zurich — 20.10.2009 P2MNet 20/54
One unique framework able to provide an effective solution to all the different IMS load-balancing issues is still lacking
One solution that integrates local, intra-domain, and inter-domain load balancing is still missing
One significantly tested solution:most papers in the IMS literature are insufficiently validated and do not include extensive experimental results collected in real-world distributed testbeds
IMS scalability:IMS scalability:open issuesopen issues
Zurich — 20.10.2009 P2MNet 21/54
IHMAS: IHMAS: emerging design guidelines emerging design guidelines for IMS scalabilityfor IMS scalability
IMS-compliant Handoff Management Application Server Active session signaling (proxy-based approach)
Intra-domain (IMS) infrastructure load balancing– Collects service-aware distributed monitor alarms– Decides and executes needed load-balancing actions
(dynamic addition/removal of CSCF components)
Intra-domain service load balancing – Adopts a data-centric session management approach to
share service state into AS pools– Exploits specific service knowledge (service awarereness)
to divide intra-domain load into partitions
Inter-domain transmission optimizations– Controls and reduces inter-domain traffic– Realizes service-aware message aggregation and batching
techniques based on distributed AS federation models
Zurich — 20.10.2009 P2MNet 22/54
IMS
DOMAIN P DOMAIN W
PS
IMS
PS
WN
W2
W1P1
P2
PN
SUBSCRIBE(to P2@domainP)
SUBSCRIBE
SUBSCRIBE
SUBSCRIBE
(to P2@domainP)
(to P2@domainP)
Inter-domain PS scenarioP: Presentity PS: Presence Server
W: Watcher
IHMAS PS scalability use caseIHMAS PS scalability use case
Zurich — 20.10.2009 P2MNet 23/54
IMS
DOMAIN P DOMAIN W
PS
IMS
PS
WN
W2
W1P1
P2
PN
Inter-domain PS scenario
PUBLISH(e.g.:”I’m online”)
NOTIFYNOTIFY
NOTIFY
NOTIFY
IMS-based components
PS is very prone to load-balancing issues!!
IHMAS PS scalability use caseIHMAS PS scalability use case
Zurich — 20.10.2009 P2MNet 24/54
IHMAS intra-domain infrastructure load-balancing– monitors distributed infrastructure and service
components (I-/P-/S-CSCFs, HSS, PS, any AS, …)
load monitoring actions tailored for the specific service (service-aware approach)
– executes application-level specific component load-balancing actions
dynamic de-/activation of distributed components and DNS (de-)registration actions
– integrates seamlessly with existing infrastructuresfull compliancy with IMS standard
P. Bellavista, A. Corradi, L. Foschini, “Enhancing the Scalability of IMS-based Presence Service for LBS Applications”, IEEE COMPSAC, 2009P. Bellavista, A. Corradi, L. Foschini, “IMS-compliant Management of Vertical Handoffs for Mobile Multimedia Session Continuity”, accepted for IEEE Communications Magazine
IHMAS intraIHMAS intra--domain PS scalability:domain PS scalability:infrastructure loadinfrastructure load--balancingbalancing
Zurich — 20.10.2009 P2MNet 25/54
Decision Maker –DM: takes load balancing and partitioning decisions
Load-Balancing Executor – LBE: enforces them
Proactive Monitoring Stub – PMSx: monitor system/ component behavior and generate overload alerts towards DM
IMS components: I-/P-/S-CSCF, HSS, PS, DNS
IHMAS intraIHMAS intra--domain PS scalability:domain PS scalability:infrastructure loadinfrastructure load--balancingbalancing
PMS: Proactive Monitoring Stub DM: Decision Maker LBE: Load-Balancing Executor
S-CSCF1
S-CSCF PMSS-CSCF
S-CSCF2
PMSS-CSCF
S-CSCFs
PS1
PS PMSPS
PSs
DB
W1 P2
P1
W2
DNS
P-CSCF1
P-CSCF PMSP-CSCF
P-CSCF1
P-CSCF PMSP-CSCF
I-CSCF
I-CSCF PMSI-CSCF
HSS
HSS PMSHSS
PS2
PMSPS
Load Balancer
IHMAS Load Balancer
DMLBE
SIP Diameter SQL SSH or UDP
Load Balancer (LB)
IHMAS Load Balancer
DMLBE
Zurich — 20.10.2009 P2MNet 26/54
Open IMS Core
IMS Bench SIPp
• IMS infrastructure
• Presence Server
• SIP traffic generation
OpenSER(now OpenSIPS and Kamailio)
Implementation hintsImplementation hints
• Deployment re-configuration Diskless Remote Boot Linux (DRBL)
Zurich — 20.10.2009 P2MNet 27/54
IMS (PS) traffic emulationIMS (PS) traffic emulation
IMS Bench SIPpIMS traffic generator that conforms to ETSI TS 186 008 IMS/NGN Performance Benchmark specificationIMS Bench SIPp permits to define benchmark configuration scenarios by composing different IMS session phases (e.g., registration, subscription, …)
Zurich — 20.10.2009 P2MNet 28/54
PS traffic benchmarkPS traffic benchmark
Three main phases (in each experiment)
1. IMS clients registration
2. Watcher subscriptions to obtain different watcher-per-presentity (w/p) ratios
3. IMS infrastructure and system stressThis phase is a mix of:
10% subscriptions, 5% registrations, 5% de-registrations, 80% publications (Presentities)
configured with various incremental steps
Zurich — 20.10.2009 P2MNet 29/54
S-CSCF
P-CSCF
IMS Bench Manager
SIPp client
IMS PS traffic
configuration andresult gathering
PresenceServer
(OpenSIPS)
S-CSCF
IMS PStraffic
SIPp client
open-ims.test
IHMAS LB
Open IMS Core
P-CSCF
32 Linux Boxes2CPU 1,8 GHz2048MB di RAM
Experimental testbedExperimental testbed
I-CSCF
S-CSCF S-CSCF
PMS PMS PMS PMS PMS
PMS PMS PMS PMS PMS
DMLBE
Intra-domainDNS
Inter-domainDNS
Zurich — 20.10.2009 P2MNet 30/54
Experimental results:Experimental results:SS--CSCF scalabilityCSCF scalability
Scalability threshold of one S-CSCF component, obtained with: 1 PS, 1 I-CSCF, 1 HSS (not shown), and 3 P-CSCFs
msec
CP
U %
0
20
40
60
80
100
120
140
160
180
200
220
0 60000 120000 180000 240000 300000 360000 420000 480000
S-CSCF PS P-CSCF1 P-CSCF2 P-CSCF3
5 cps 10 cps 15 cps 20 cps 25 cps 30 cps 35 cps 40 cps 45 cps 50 cps 55 cps 60 cps 65 cps 70 cps
Zurich — 20.10.2009 P2MNet 31/54
Experimental results:Experimental results:dynamic Sdynamic S--CSCFsCSCFs loadload--balancingbalancing
0
20
40
60
80
100
120
140
160
180
200
220
0 60000 120000 180000 240000 300000 360000 420000 480000
S-CSCF1 S-CSCF2
5 cps 10 cps 15 cps 20 cps 25 cps 30 cps 35 cps 40 cps 45 cps 50 cps 55 cps 60 cps 65 cps 70 cps
msec
CP
U %
Triggered by the combination of a NOTIFY alarms
Filtering and prediction techniques permits to effectively smooth sporadic peaks
DM activates S-CSCF2
PMSPS alarmPMSS-CSCF1 alarm
Zurich — 20.10.2009 P2MNet 32/54
IHMAS intra-domain service load-balancing– extends IMS PS to support multiple AS service state
storages and fast exchange of (and access to) shared session state among ASs
novel PS intra-domain module to enable data distribution overlays and caching techniques within AS partition
– exploits existing standards for data distributiondata distribution is fully compliant with Data
Distribution Service (DDS), an Object Management Group (OMG) standard
– divides intra-domain service workload by applying a divide-and-conquer principle (for big domains)
IMS routing based on HSS and IMS filter criteriaP. Bellavista, A. Corradi, L. Foschini, “Understanding and Enhancing the Scalability of IMS-based Services for Wireless Local Networks”, IEEE WLN, 2009
IHMAS intraIHMAS intra--domain PS scalability:domain PS scalability:service load balancingservice load balancing
Zurich — 20.10.2009 P2MNet 33/54
DDS
IHMAS intraIHMAS intra--domain PS scalability:domain PS scalability:session datasession data--centric managementcentric management
IMS
IMS
• Service state (both subscriptions and publications) stored locally at PS DB
• PUBLISH write-through (using DDS)
Zurich — 20.10.2009 P2MNet 34/54
Message routing
Statistical load partition based on a hash function evaluated on presentity identifiers exploits HSS + IMS filtering criteria
IHMAS intraIHMAS intra--domain PS scalability:domain PS scalability:static balancing among PS poolsstatic balancing among PS pools
Zurich — 20.10.2009 P2MNet 35/54
IHMAS inter-domain service optimizations– extends IMS PS to support message aggregation/batching
(diminishes the number of inter-domain NOTIFY transmissions)
novel PS inter-domain optimization module for NOTIFY message parsing and inter-domain routing
– supports mobile clients and service differentiation (gold, silver, copper, …)
Gold: instant presence info delivery high costSilver: slightly delayed presence info delivery medium costCopper: very delayed presence info delivery low cost
– integrates seamlessly with existing infrastructuresfull compliance with IMS standard
P. Bellavista, A. Corradi, L. Foschini, “IMS-based Presence Service with Enhanced Scalability and Guaranteed QoS for Inter-Domain Enterprise Mobility”, IEEE Wireless Communications Magazine, vol. 16, no.3, Jun. 2009
IHMAS interIHMAS inter--domain PS scalability:domain PS scalability:transmission optimizationstransmission optimizations
Zurich — 20.10.2009 P2MNet 36/54
Common NOTIFYCommon NOTIFY
IMS
PS
IMS
PS
WN
SUBSCRIBE
SUBSCRIBE
SUBSCRIBE
(to P2@domainP)
W2
W1
(to P2@domainP)
(to P2@domainP)SUBSCRIBE
P2
“Several watchers subscribed to one presentity”
Zurich — 20.10.2009 P2MNet 37/54
IMS
PS
IMS
PS
WN
NOTIFY
NOTIFY
NOTIFY
W2
W1
P2
PUBLISH
NOTIFY +
AggregatesNOTIFY messages
at Presentity’sdomain
1 only inter-domain NOTIFYmessage
NOTIFY messagescreation at
Watchers’s domain
Watchers’list
Common NOTIFYCommon NOTIFY
Zurich — 20.10.2009 P2MNet 38/54
IMS
PS
IMS
PS SUBSCRIBE(to P1@domainP,
P2@domainP,PN@domainP)
W2
SUBSCRIBE
P2
PN
P1
Batched NOTIFYBatched NOTIFY
“One single watcher subscribed for multiple presentities”
Zurich — 20.10.2009 P2MNet 39/54
Batched NOTIFYBatched NOTIFY
Time-based (periodic) NOTIFY message batching
IMS
PS
IMS
PSW2
P2
PN
P1
PUBLISH
PUBLISH
PUBLISH
NOTIFY
NOTIFY
NOTIFY
Zurich — 20.10.2009 P2MNet 40/54
Batched NOTIFYBatched NOTIFY
IMS
PS
IMS
PSW2
P2
PN
P1
PUBLISH
PUBLISH
PUBLISH
NOTIFY
NOTIFY
NOTIFY
NOTIFY
only 1 inter-domainNOTIFY message
NOTIFY
Zurich — 20.10.2009 P2MNet 41/54
WA.4 WA.1 P-/S-CSCFA IOMA + PSAWA.3 I-/S-CSCFB IOMB + PSB PB.1 PB.4
DomainA DomainB
(1) SUBSCRIBE (2) SUBSCRIBE (3) SUBSCRIBE (4) SUBSCRIBE
(5) 200 OK(6) 200 OK(7) 200 OK(8) 200 OK(9) PUBLISH
(10) OK
SUBSCRIBEfrom other watchers
(11’) Common NOTIFY(one per presentity)
PB.4
PUBLISHfrom other presentities
(11) NOTIFYGo
ldC
lien
ts
(13’) NOTIFY msgs
(12) NOTIFY
(14’) NOTIFY
(13) OK (14) OK
(12’) OK
(15’) OK msgs (16’) OK
(11’’) Batched NOTIFY(one per watcher)
(13’’) aggregated NOTIFY(14’’) NOTIFY
(12’’) OK
(15’’) OK msgs (16’’) OK
Silv
erC
lient
sC
oppe
rC
lient
s
PrioritizedNOTIFY schedule
Watchers Authorization andNOTIFY construction
Watcher Authorization, NOTIFY aggregation, NOTIFY construction
WA.4WA.4 WA.1WA.1 P-/S-CSCFAP-/S-CSCFA IOMA + PSAIOMA + PSAWA.3WA.3 I-/S-CSCFBI-/S-CSCFB IOMB + PSBIOMB + PSB PB.1PB.1 PB.4PB.4
DomainA DomainB
(1) SUBSCRIBE (2) SUBSCRIBE (3) SUBSCRIBE (4) SUBSCRIBE
(5) 200 OK(5) 200 OK(6) 200 OK(6) 200 OK(7) 200 OK(7) 200 OK(8) 200 OK(8) 200 OK(9) PUBLISH(9) PUBLISH
(10) OK(10) OK
SUBSCRIBEfrom other watchers
(11’) Common NOTIFY(one per presentity)
PB.4PB.4
PUBLISHfrom other presentities
(11) NOTIFYGo
ldC
lien
ts
(13’) NOTIFY msgs
(12) NOTIFY
(14’) NOTIFY
(13) OK(13) OK (14) OK(14) OK
(12’) OK(12’) OK
(15’) OK msgs(15’) OK msgs (16’) OK(16’) OK
(11’’) Batched NOTIFY(one per watcher)
(13’’) aggregated NOTIFY(14’’) NOTIFY
(12’’) OK(12’’) OK
(15’’) OK msgs(15’’) OK msgs (16’’) OK(16’’) OK
Silv
erC
lient
sC
oppe
rC
lient
s
PrioritizedNOTIFY schedule
Watchers Authorization andNOTIFY construction
Watcher Authorization, NOTIFY aggregation, NOTIFY construction
IHMAS interIHMAS inter--domain PS scalability:domain PS scalability:PS protocol enhancementsPS protocol enhancements
Experimental results: CPU load at border CSCF components
Zurich — 20.10.2009 P2MNet 42/54
0
5
10
15
20
25
8
20 33 46 58 71 84 96
109
122
135
147
160
173
186
198
211
224
237
250
263
276
289
302
315
328
341
354
10 CPS5 CPS 15 CPS 20 CPS 25 CPS 35 CPS 45 CPS 55 CPS 65 CPS 70 CPS 80 CPS 90 CPS
0
5
10
15
20
25
8
20 33 46 58 71 84 96
109
122
135
147
160
173
186
198
211
224
237
250
263
276
289
302
315
328
341
354
10 CPS5 CPS 15 CPS 20 CPS 25 CPS 35 CPS 45 CPS 55 CPS 65 CPS 70 CPS 80 CPS 90 CPS
0
25
50
75
100
125
150
175
200
225
8 20 33 46 58 71 84 96 109
122
135
147
160
173
186
198
211
224
237
250
263
276
289
302
315
328
341
354
15 CPS 25 CPS 45 CPS 65 CPS 70 CPS5 CPS 90 CPS80 CPS10 CPS 20 CPS 35 CPS 55 CPS
6.6 w/p 8.6 w/p 10.9 w/p 14.6 w/p
sec
CP
U %
ME
MO
RY
%
sec
Experimental results:Experimental results:w/out IHMAS optimizationsw/out IHMAS optimizations
CPS: Calls Per Secondw/p: mean number watcher subscriptions for each presentity
Zurich — 20.10.2009 P2MNet 43/54
0
10
20
30
40
50
60
70
80
90
8
21 34 46 59 72 84 97 110
122
135
148
160
173
186
198
211
224
237
249
262
275
287
300
313
325
338
351
15 CPS 25 CPS 45 CPS 65 CPS 70 CPS5 CPS 90 CPS80 CPS10 CPS 20 CPS 35 CPS 55 CPS
0
10
20
30
40
50
60
70
80
90
8
21 34 46 59 72 84 97 110
122
135
148
160
173
186
198
211
224
237
249
262
275
287
300
313
325
338
351
15 CPS 25 CPS 45 CPS 65 CPS 70 CPS5 CPS 90 CPS80 CPS10 CPS 20 CPS 35 CPS 55 CPS
0
2
4
6
8
10
12
14
8 21 34 46 59 72 84 97 110
122
135
148
160
173
186
198
211
224
237
249
262
275
287
300
313
325
338
351
15 CPS 25 CPS 45 CPS 65 CPS 70 CPS5 CPS 90 CPS80 CPS10 CPS 20 CPS 35 CPS 55 CPS
Experimental results:Experimental results:with IHMAS common NOTIFYwith IHMAS common NOTIFY
6.6 w/p 8.6 w/p 10.9 w/p 14.6 w/p
sec
CP
U %
ME
MO
RY
%
sec
CPS: Calls Per Secondw/p: mean number watcher subscriptions for each presentity
Zurich — 20.10.2009 P2MNet 44/54
0
10
20
30
40
50
60
70
6
18 31 44 56 69 82 95 107
120
133
145
158
171
183
196
209
222
234
247
260
273
285
298
311
323
336
349
15 CPS 25 CPS 45 CPS 65 CPS 70 CPS5 CPS 80 CPS10 CPS 20 CPS 35 CPS 55 CPS 90 CPS
0
10
20
30
40
50
60
70
6
18 31 44 56 69 82 95 107
120
133
145
158
171
183
196
209
222
234
247
260
273
285
298
311
323
336
349
15 CPS 25 CPS 45 CPS 65 CPS 70 CPS5 CPS 80 CPS10 CPS 20 CPS 35 CPS 55 CPS 90 CPS
0
1
2
3
4
5
6
7
8
6
18 31 44 56 69 82 95
107
120
133
145
158
171
183
196
209
222
234
247
260
273
285
298
311
323
336
349
15 CPS 25 CPS 45 CPS 65 CPS 70 CPS5 CPS 80 CPS10 CPS 20 CPS 35 CPS 55 CPS 90 CPS
0
1
2
3
4
5
6
7
8
6
18 31 44 56 69 82 95
107
120
133
145
158
171
183
196
209
222
234
247
260
273
285
298
311
323
336
349
15 CPS 25 CPS 45 CPS 65 CPS 70 CPS5 CPS 80 CPS10 CPS 20 CPS 35 CPS 55 CPS 90 CPS
Experimental results:Experimental results:with IHMAS batched NOTIFYwith IHMAS batched NOTIFY
6.6 w/p 8.6 w/p 10.9 w/p 14.6 w/p
sec
CP
U %
ME
MO
RY
%
sec
CPS: Calls Per Secondw/p: mean number watcher subscriptions for each presentity
Zurich — 20.10.2009 P2MNet 45/54
Experimental results:Experimental results:number of internumber of inter--domain domain NOTIFYsNOTIFYs
1
10
100
1000
10000
100000
1000000
6.6 8.6 10.9 14.6 w/p
# N
OT
IFY
me
ss
ag
es
w/o PS enhancements w/ Common NOTIFY w/ Batched NOTIFY
Zurich — 20.10.2009 P2MNet 46/54
Experimental results:Experimental results:interinter--domain NOTIFY delaydomain NOTIFY delay
1
10
100
1000
6.6 8.6 10.9 14.6 w/p
ms
w/o service differentiation gold client silver client
NO
TIF
Y m
essa
ge
del
ay
Zurich — 20.10.2009 P2MNet 47/54
ConclusionsConclusions
Strong need for IMS scalable solutions– Both at the infrastructure and service level– Context- and service-aware approaches seem
to be promising and should not be neglected
Interoperability and standard compliancy– Full IMS standard compliance for
inter-domain optimization techniques – Ad-hoc solutions and integration with other
emerging standards at intra-domain level
Real-world testbeds should be employed whenever possible
Zurich — 20.10.2009 P2MNet 48/54
Research directionsResearch directions
Context-aware and self-* middleware solutions for service state management– Scalability is a complex and still open task
Session state grain/footprint/dissemination– Standard (r)evolution? (SIP, IMS, …)
Session control !!!!– Scalability first (millions/billions of nodes, systems of systems,…)
– Use and interaction with different standards Example: OMG Data Distribution Service (DDS) to ease
and boost context and presence data dissemination
And several others…
Zurich — 20.10.2009 P2MNet 49/54
Thank youThank you
Zurich — 20.10.2009 P2MNet 50/54
Many thanks go also toMany thanks go also to……
My research group and especially the B.Sc. and M.Sc. students who decided to collaborate with me on the IHMAS project– S. Parcaroli, G. Carella, R. Colombari, C. Pitscheider, L.
Scalorbi, F. Checchi, D. Campellone, and L. Nardelli
Professors A. Boukerche, S. Samarah, L. Mokdad, and A.-E. M. Taha for asking me to give this keynote speech
Zurich — 20.10.2009 P2MNet 51/54
IHMASIHMASproject web site and contacts project web site and contacts
Prototype code: http://lia.deis.unibo.it/Research/IHMAS
Contacts: Luca Foschini ([email protected])
Thanks for your attention!
Zurich — 20.10.2009 P2MNet 52/54
Some relevant references Some relevant references
G. Camarillo and M.A. García-Martín , The 3G IP Multimedia Subsystem (IMS) – Second Edition, Wiley, 2006.A. Dutta et al., “Mobility Testbed for 3GPP2-Based Multimedia Domain Networks”, IEEE Communications Magazine, vol. 45, no. 7, Jul. 2007.L. Skorin-Kapov et al., “Application-Level QoS Negotiation and Signaling for Advanced Multimedia Services in the IMS”, IEEE Communications Magazine, vol. 45, no. 7, Jul. 2007.T.T. Kwon et al., “Mobility management for VoIP service: Mobile IP vs. SIP”, IEEE Wireless Communications, vol. 9, no. 5, Oct. 2002.M. Stemm and R.H. Katz., “Vertical Handoff in Wireless Overlay Networks”, Mobile Networks and Applications, vol. 3, no. 4, Kluwer, Jan. 1998.M.E. Kounavise et al., “Design, implementation, and evaluation of programmable handoff in mobile networks”, Mobile Networks and Applications, vol. 6, no. 5, Kluwer, Sept. 2001.
Zurich — 20.10.2009 P2MNet 53/54
Some relevant references Some relevant references
V. Bharghavan et al., “The TIMELY adaptive resource management architecture”, IEEE Personal Communications, vol. 5, no. 4, Aug. 1998.H. Schulzrinne and E. Wedlund, “Application-layer mobility using SIP”, ACM Mobile Computing and Communications Review, vol. 4, no. 3, Jul. 2000.D. Vali et al., “An efficient micro-mobility solution for SIP networks”, IEEE GLOBECOM, 2003.S.K. Das, “SIP-based vertical handoff between WWANs and WLANs”, IEEE Wireless Communications, vol. 12, no. 3, Jun. 2005.N. Banerjee et al., “SIP-based Mobility Architecture for Next Generation Wireless Networks”, IEEE PerCom, 2005.C. Kalmanek et al., “A Network-Based Architecture for Seamless Mobility Services”, IEEE Communications Magazine, vol. 44, no. 6, Jun. 2006.A. Udugama et al., “NetCAPE: Enabling Seamless IMS Service Delivery across Heterogeneous Mobile Networks”, IEEE Communications Magazine, vol. 45, no. 7, Jul. 2007.
Zurich — 20.10.2009 P2MNet 54/54
Some of our recent contributionsSome of our recent contributions
P. Bellavista , A. Corradi, L. Foschini, “Context-Aware Handoff Middleware for Transparent Service Continuity in Wireless Networks”, Pervasive and Mobile Computing Journal, Elsevier Science, vol. 3, no. 4, Aug. 2007.P. Bellavista, A. Corradi, L. Foschini, “IMS-based Presence Service
with Enhanced Scalability and Guaranteed QoS for Inter-Domain Enterprise Mobility”, IEEE Wireless Communications Magazine, SI on Enterprise Mobility Services, vol. 16, no.3, Jun. 2009.P. Bellavista, A. Corradi, L. Foschini, “Enhancing the Scalability of
IMS-based Presence Service for LBS Applications”, IEEE COMPSAC, 2009.P. Bellavista, A. Corradi, L. Foschini, “Understanding and Enhancing the Scalability of IMS-based Services for Wireless Local Networks”, to appear as invited paper in the Proc. of IEEE WLN'09, held in conjunction with the IEEE LCN, 2009.P. Bellavista, M. Cinque, D. Cotroneo, L. Foschini, “Self-Adaptive Handoff Management for Mobile Streaming Continuity”, to appear in IEEE Transactions on Network and Service Management.