information centric networks - the next...
TRANSCRIPT
IINFORMATION CENTRIC NETWORKS --THE NEXT GENERATION INTERNET
Anantha Simha
Principal Scientist & HeadNetworks Lab, TCS Research, Bangalore
2Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
OutlineCurrent Internet Architecture
Evolution of InternetChallenges in the Current Internet
ICN – the Emerging Networking AlternativeRealizing ICNICN and SDN – Internet MatrimonyICN – Representative ArchitecturesICN – BenefitsICN Research ChallengesDiscussions
3Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Internet is EvolvingNot possible to define life without Internet
No more a luxury; a must have itemNot only connecting people, it is connecting the things…
Most of the Evolution of Internet happened in the last decadeDeveloped and deployed in 1960s and 1970s…Post Y2K, new era of Internet started
Transparent Network LayerHour-Glass waist
Internet in the past Internet Today Internet in the Future
4Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Challenges in the Current InternetChanges in the Trend
Re-defined applicationsEvolution of social media, online gaming, video on demand, etc.
Data creation, consumption and transferMassive - 2.5 Exabytes of data per day !! and growing…
• >70% of the current traffic belongs to YouTube, Netflix, Amazon, iTunes, Facebook…• In every 5 minutes of real-time 60 hours of video is uploaded to YouTube
Network CongestionCore network
Limited bandwidth Access network
Capacity is also a major concernLocal vs. global copies of the data
Multiple copies transmission• Major problem with mobile networks
Source: Cisco Visual Networking Index: Forecast and Methodology, 2010-2015
5Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Challenges in the Current InternetOther Concerns
Network capabilityMulticast, broadcast are not exploited fully
Security is end-to-endAttempting to secure the pipe, instead of the content
Reliability for content accessDifficult in case of network failureSearch engine should be accessible directly
Any other Thought?Shouldn’t we concentrate on the content, not on the host?
In both disruptive and non-disruptive linksIn providing content security rather than pipe securityClean slate or overlay solution?Transformation or Translation ??
Information Centric Networks[ICN] is an attempt to answer these questions !!
6Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Current Internet Architecture
Internet Service ProviderUsers/Subscribers
Servers/Sources/Storage/Publishers
Search EngineDNS
Data Communication is between ‘host-to-host’ Source and user addresses are to be uniqueDNS resolution and legacy routing are the keyUses TCP/IP Protocol extensively
Routers/Gateways
7Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Information Centric Networks (ICN)Information or the Content is the key
Information is decoupled from the sourceIndependent of location, application, storage, transportation, caching etc.Host and connections could be untrusted
Network manages the information, not the end-points (sources)Network should be aware of the information and its current location
All existing applications must be supported
REQUEST
PUBLISH
Networks become a Black BoxUsers Servers/Sources/Storage
Search EngineCache
Cache
Cache
Router/Cache CacheH
8Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Requirements of ICNThree Parties should be Happy
UserImproved QoS, Transparent systemMultiple devices should send/receive information
OperatorConnecting to diverse networks – wired, wireless Self-organize using any available resource, device, or channelBackward compatibilityReduced CAPEX and OPEX
ProducerPush content while anticipating and inferring users’ need/demandData should be accessible
Should Support the Regulators as wellRegulation and Legal issues become crucial
9Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Features of ICNFocus on information/content - NDO
Web pages, videos, photos etc.Location independentCopies are equivalentGranularity varies – packet to full contentMeta data associated – publisher, version etc.Information is a network layer function
Focus on NamingInformation is named, matched and addressedCan be located anywhere, location – identity splitRetrieval is user driven
Focus on Information DeliveryInternet is not content aware - application layer workarounds[web caching]Can handle surges for content unlike overlays like CDNsSame information should not travel more than once at a time on the same link
Allows aggregation of requests for multicast delivery
10Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Features of ICNFocus on Mobility
Internet addressing scheme designed for fixed hostsMobility is a key issue - Mobile IP patch is a temporary solution
In ICN, Host mobility is addressed by Publish/Subscribe paradigmBrokers provide rendezvousPublishers/Subscribers are decoupled in space/time – enables mobility
Focus on SecurityInternet operates in trusted environment
User and data auth/privacy was not a requirementAllows DoS attacks Bad data still gets forwarded even with patches like FW, IPSec etc.
In ICN data is interest driven Reduces SpamSelf certifying names allows in-network malicious data filteringDecoupling prevents DoS like attacks
11Experience certainty.
ICN Timeline
DONA – Data Oriented Network ArchitectureNDN – Named Data Networking
Source – A survey of ICN ResearchCOMMUNICATIONS SURVEYS AND TUTORIALS, VOL. 16, NO. 2, 2014, PP. 1024–1049;
12Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Realizing ICNArchitectureFunctions of Network Components
Naming, Name resolutionRouters, gateways, cache routers, etc
Security of the DataCurrent Internet vs. ICN
Extending ICN to Wireless Networks
A natural fit
Operator NetworkUsers
Sources
Search EngineCache
Cache
Cache
Cache/Router/Gateway
Access Network
13Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Architecting the ICNPossible Solutions
Tweaking the current Internet architecturePatches are only temporal solution
• Current Internet with whatever modifications may not address all the issues
Clean-slate solution Good, but has many deployment and business challenges
• Should we hamper the existing Internet setup?• Should we change the TCP/IP protocol stack?
Incremental Solution can be an alternativeOverlay network architecture can be beneficial
Must ensure backward compatibility• User as well as network friendly• Business and deployment interests will be there
Still open for Research
1
14Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Overlay ICN Architecture !!1
What do we want?Both the Legacy Networks and the ICN should co-exist
Incremental design to support backward compatibilityShould be transparent to the userCost of deployment and Operation should be Minimum
QoS should be ImprovedO-ICN Architecture
Three Major ComponentsICN User and Source
Should identify the requestsICN Router
Should be capable of cachingICN Manager
Control plane activities are performed
15Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Naming in ICN2
Most important in ICNInformation objects have location independent unique IDsAll object copies share same ID
Security applies to NameName-data integrity – bindingObject authenticity – correct content Provenance – author, date of creation, ver, etc.
Flat vs. HierarchicalNames are assigned as sequence of symbols with no structure No clear relationship between any two namesSelf certifying, Unique, Persistent Searching takes more time - O(n2)
Ex - ni://tcd.ie/sha256:NDVmZTMz
16Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
tcs.com
research
paper.pdf
_v1
_s1
_v2
_s0
HierarchicalNames are sequence of symbols with specific structure - URLs
Follows Parent/child semantics Human readableScalable
Searching is easyRooted to a publisher prefix
/tcs.com/research/paper.pdf/_v2/_s0
User supplied nameVersion
Chunk
Naming in ICN2
17Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Routing in ICNLocates a data object based on its name
Publish stepEstablishes path to source
Name resolution stepTranslates name into its locator/sourceSimilar to DNS
Discovery stepRoutes request to sourceCentralized/distributed
Delivery stepRoutes content/object back to the requesterCoupled/decoupled with name resolutionIP based or non-IP basedCached en-route
18Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Caching in ICNWhy Caching?
Reduces redundant transmission of same contentICN routers cache contents flowing through themCan serve locally for any future demand of the content
Where all to cache?Caching on all or few nodes/routers?Size of cache
Caching policy Goal
Min delay and max cache hitMin processing overhead
Which policy?LRU, FIFO, etc. ,are too genericOn-demand caching Vs proactive caching Centralized vs. distributed caching
Consumer
Cached Routers
Publisher
3
Cache Miss
?
Consumer
Cache Hit
Publisher
19Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
RequestsContents
On-path
Off-path
Caching Types and ChallengesBroad types
On-path caching: caching on data pathLess network intervention Which routers/nodes to be cached
• Depends on traffic characteristics
Off-path caching: other than the return pathShould be cached at strategic locationsLess duplications if off-path
Managing caches How to redirect requests to nearest cachesCache update
Existing version may have updated at original source
Cache policy for ICN is still not explored much !!Proposed Universal Caching
Optimal caching policyCurrent research on distributed and collaborative caching
B3
20Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
Security in ICNSecurity in the Internet
End-to-End SecurityPipe is secured, not the individual content
Rigid Protocol Stack ArchitectureIntegrating accountability mechanism is a challengeNo cross layer information sharedEnd to End security is still vulnerable
• DoS and Distributed DoS (DDoS) attacks
ICN is interest driven network Need basis data/content/packet is transferred
Source does not matterICN security relies upon trusted naming verification
• Owner names the content through efficient hashing, generates pub/prv key• Receiver verifies content is indeed signed by owner• Evidence based trustworthiness of publisher• Efficient/trust worthy mapping service required
4
21Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
ICN in Wireless NetworksDecoupling the Content from the Producers
Good for disruption toleranceCan be managed at the operator level, locally
Storage a part of Network ServiceLocal routers can store the contents
Eg. – Train, Bus, Flights, movies, songs, games, etc
IoT Solutions can also be SupportedAccess Networks
Simultaneous usage of the Access NetworksWiFi, Bluetooth, Cellular networks Multiple paths to improve QoS
Reduced demandQoS Improvement
6
22Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
ICN in Mobile Network – Future !!Boon to 5G Networks!!Consumer mobility vs. Producer mobility
Mobile user can be a data consumer and/or a data producer as wellDownloading movies or watching movies vs. video/audio conferencing
Re-registration is a major challengeConsumer/User Mobility
• Issues another Request/interestSource/Producer Mobility
• Challenge, requires Name register updates in some cases
User
Source
Source
Moving Source
SourceMoving User
6
23Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
ICN in Mobile Network – RealizationAt Application Layer
Scoping of Content and LocationRemoval of Connection Oriented Sessions
May be some other techniques can be usedResilience through Replication
In-route caching/content multi-homingCreation of communities/groups to handle the mobility
Virtualization of the RANUse of centralized resource management technique
Dynamic allocation of BTSs/eNBs to the consumers or producers
Internet
Virtualized RANSDN Controller
OpenFlow Enabled Network
6
24Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
ICN and SDN – Internet Matrimony !!!World is moving towards Software Defined *.*
SDN a necessity & reality of the core and access network
Synergy between ICN supporting efficient data dissemination and SDN providing flexible management is the need of the future
Any architecture for ICN should support SDN Technology
ICN talks about WHAT and SDN talks about HOW that is to be communicated
ICN talks about separating function of information demand and supplySDN talks about separation of network function
The HOW of ICN can be a part of SDN Controller !!
O-ICN Manager and SDN Controller can Co-exist
COMET and O-ICN Support SDN
25Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
SoA - Data Oriented Network Architecture (1/3) It is the first complete implementation (2007) of ICN architectureNaming: Flat names instead of Hierarchical URLs
Self Certifying, application/location independent, globally unique naming
Name ResolutionNames are published/pushedOverlay name resolution mechanismSpecialized servers - Resolution Handlers (RHs)/Name Resolvers (NR)
At least one RH per AS (Autonomous System)RHs are interconnected, forming a hierarchical name resolution service on top of inter-domain routing
RoutingRouting can be coupled or decoupled from name
Maintains IP addressing locally (de-coupled routing) or globally (coupled routing)
26Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
SoA - Data Oriented Network Architecture (2/3)
NR/RH
Tier-1 AS
NR/RH
AS3
NR/RH
AS2
NR/RH
AS1
UserSource
Peering Link
Client-Provider LinkRegister Message(1-3)
Data(8-11)Find Message(4-7)
NR: Name ResolverRH: Resolution HandlerAS: Autonomous System
Names – Flat, form –P:LResolution – Through NR/RH infraRouting – Coupled or decoupledCaching – on path, via RHMobility – user to issue new Find,
while source must re-registerSecurity – Self CertifyingMulticasting – Through Findaggregation
UC Berkley Proposal2007
27Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
State of The Art – DONA (3/3) Caching
On –path caching via RH infrastructureMobility
Users need to issue new FIND messagesSource (Publishers) need to re-register
SecuritySelf certifying
MulticastingBy caching FIND messages in RHs for a specific period of time and sending information till they expire
Additional FIND messages are merged at the RH such that multicast distribution tree can be formed with the help of reverse routing
28Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
SoA – Named Data Networking (1/3) Another full-fledged ICN architecture
Termed as Named Data Networks or Content Centric NetworksBasic idea was published in Google Tech Talk (2006)
Names in NDNHierarchical names – name aggregation possibleSimilar to URLs, but not necessarily URLs
Information can be segmented, user’s application should be aware of it.Users can ask for dynamic (future) contents
Name Resolution and RoutingName resolution and routing are coupled - OSPF can be usedSubscribers issue INTEREST messages to request information objects Information is arrived in the form of DATA messages
Symmetric routing for INTEREST and DATAMessages are forwarded hop-by-hop by Content Routers (CRs)
29Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
FIB
Name Next
/aueb.gr/ CR C
/aueb.gr/cs CR B
PITName Requested
/aueb.gr/ai/new.htm User
CSName Data
- -
A’s tables after receiving Interest packet
FIB
Name Next
/aueb.gr/ CR C
/aueb.gr/cs CR B
PITName Requested
- -
CSName Data
/aueb.gr/ai/new.htm …
A’s tables after receiving Data packet
FIB
Name Next
/aueb.gr/ Source1
PITName Requested
/aueb.gr/ai/new.htm CR A
CSName Data
- -
C’s routing tables after receiving Interest packet
FIB
Name Next
/aueb.gr/ Source1
PITName Requested
- -
CSName Data
/aueb.gr/ai/new.htm …
C’s routing tables after receiving Data packet
LinkInterest Message(1-3)Data(4-6)
User
Source1
Source2
CR A
CR B
CR C
FIB
Name Next
/aueb.gr/cs Source2
PITName Requested
- -
CSName Data
- -
State of The Art – NDN (2/3)
CR: Content Router
30Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
State of The Art – NDN (3/3) Caching
Supports on-path and off-path cachingEach CR first consults its CS whenever it receives an INTEREST message
• It caches all the DATA messages for the same INTERESTCS uses LRU in general
Strategy layer can direct the INTEREST to a CDN server rather than to the originating publisher
SecuritySecurity layer added to provide security to the dataEach DATA message is secured through a signature over the name and the information content
MobilityUser mobility – issue new INTEREST
Both routes will be having the dataSource mobility – FIB needs to be updated
Uses LBFL – Listen First Broadcast Later protocol
31Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
ICN Internet – Benefits !!!Efficiency
Improved network usage and efficiencyBandwidth optimization
ScalabilityImproved scalability with respect to information/bandwidth demand
RobustnessImproved robustness in challenging communication scenario
SecurityImproved security of the content, through name-data integrity
Mobility & multi-homingPublish/Subscribe modelNon real time
Reduced OPEX and CAPEX!!
32Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
ICN Research ChallengesProvider Mobility
Challenge as it is necessary to update locator info# of objects exceeds # of hosts
Caching mitigates but selectivelyCauses convergence delays and stale info
Consumer MobilityRe-registration takes finite timePotential to get no/stale data
Real-time handoff delaysMobility during real time communication is a big issue
Has little potential for cachingAll nodes are both producers and consumersCentralized resolution systems may perform better
33Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
ICN Research ChallengesHuge Data Set
Each data set/content/information is separately named, resulting in naming, name resolution, routing and caching problems
Privacy & SecurityRequires nodes to exposes their interests to networkThird parties can study this pattern and undermine privacySecurity is compromised as rogue nodes can publish and compromise routing
Practical deploymentOne that exceeds all othersBenefits will accrue only entire network is ICN readyDiversity of ICN proposals make this near impossibleCould lead to islands of improvement
34Experience certainty.
Copyright 2015 | Tata Consultancy Services Limited | Strictly Confidential
ReferencesG. Xylomenos et.al., "A Survey of Information-Centric Networking Research", IEEE Communications Surveys & Tutorials, Vol. PP, No. 99. (2013), pp. 1-26,B. Ahlgren et.al., “A survey of information-centric networking,” Communications Magazine, IEEE, vol. 50, pp. 26–36, July 2012.D. Cheriton et.al., “TRIAD: A New Next-Generation Internet Architecture,” 2000.V. Jacobson, et.al., “Networking named content,” in Proc. of the 5th international conference on Emerging networking experiments and tech., pp. 1–12, ACM, 2009.T. Koponen, et.al., “A data-oriented (and beyond) network architecture,” in ACM SIGCOMM Computer Communication Review, vol. 37, pp. 181–192, ACM, 2007.N. McKeown, “Software-defined networking,” INFOCOM keynote talk, 2009. S. Shenker, et.al., “The future of networking, and the past of protocols,” Open Networking Summit, 2011.“Publish Subscribe Internet Technology (PURSUIT).” http://www.fp7pursuit.eu/PursuitWeb/.A. Detti, et.al., “Comet: a content centric inter-networking architecture,” in Proc. of the ACM SIGCOMM workshop on Information-centric networking, pp. 50–55, ACM, 2011.S. Shailendra, B. Panigrahi, H. K. Rath, and A. Simha, “A novel architecture for information centric networking (icn),” in IEEE National Conference on Communications (NCC), Mumbai, India, Feb-March 2015.B. Panigrahi, S. Shailendra, H. K. Rath, and A. Simha, “Universal caching model and markov-based cache analysis for information centric networks,” in ANTS, 2014 IEEE International Conference, New Delhi, India, 2014.