Local ocal-Aware ware IC ICN-Based Based Co Cont ntent ent - - PowerPoint PPT Presentation

local ocal aware ware ic icn based based co cont ntent ent
SMART_READER_LITE
LIVE PREVIEW

Local ocal-Aware ware IC ICN-Based Based Co Cont ntent ent - - PowerPoint PPT Presentation

4/2016 o, 11/04/2016 ncisco, , San Francisc om, ocom p, Infoc hop, ksho NOM Works Eco conomic nomic In Incen centi tive ves s fo for r Dep eployin loying Local ocal-Aware ware IC ICN-Based Based Co Cont ntent ent Deli


slide-1
SLIDE 1

Eco conomic nomic In Incen centi tive ves s fo for r Dep eployin loying Local

  • cal-Aware

ware IC ICN-Based Based Co Cont ntent ent Deli eliver very

Patrick Truong, Bertrand Mathieu Orange Labs NOM Works ksho hop, p, Infoc

  • com
  • m,

, San Francisc ncisco,

  • , 11/04/2016

4/2016

slide-2
SLIDE 2

2

Timeline

  • Evolution of business models and actors relationship with

ICN deployment

  • Local-aware NDN routing architecture

– use-case : Online Social Network (OSN) applications

  • Economic Incentives for the local-aware ICN-based content

delivery

  • Conclusion
slide-3
SLIDE 3

3

  • Current ecosystem: well installed & identified players
  • End-users (Eyeballs)
  • consume contents provided by CP
  • Content Publisher:
  • Provides content (UGC or premium) on Internet over a given

geographical area (possibly defined according to distribution rights)

  • CDN providers (can be CP: e.g.; GC)

– Manages optimized content storage and delivery resources to sell scalable and quality content storage and delivery services

  • Transit ISP

– Provides transport services to CPs and access ISPs to reach Internet

  • Access ISP

– Provides Internet Access service to end-users (eyeballs)

Evolution of business models and actors relationship with ICN deployment (1/)

( )

slide-4
SLIDE 4

ICN Main features

 user

r mobi bilit lity

 multipat

path

 multicast

astin ing

 caching

ng

 content

ent prote tect ction ion and authenticat ication ion

 => Can change the delivery chain and business models

(“transparent” caching, “transparent” multicasting)

3G WiFi

slide-5
SLIDE 5

5

Evolution of business models and actors relationship with ICN deployment (1/)

  • Main consideration : Keep Actors and values for them, from a smooth integration

and deployment of ICN, valubale for each of them

  • => no strong opposition
  • ICN-based content delivery, provided by CDN providers or ISP providers or

both

ICN IP

Level of contractualisation between ISP and 3rd parties Business Model 1 Business Model 3 Business Model 4

Contracted CPs services Non- Contracted CPs services ISP own services

slide-6
SLIDE 6

6

Local-aware NDN routing architecture (1/)

  • We consider Twitter for the sake of clarity
  • The networking behaviour is not in line with the end-users’

behaviour

– Users are very frequently connected to other users in the same town

  • r region, in short in the immediate vicinity. A majority of tweets are

destined to local or close users, except for very popular accounts. – But almost all the network traffic is send to the remote servers (US Twitter servers or US/Ireland Facebook server) and all consumed tweets are retrieved from the remote servers

  • What can we do to optimize the delivery of contents, more in line

with local users relationships ?

– NDN (Named Data Networking): a candidate for improvement

slide-7
SLIDE 7

7

NDN-based Routing for Local-Awareness in Twitter-like Social Networks

  • We use the NDN architecture to optimize the networking behaviour while better

reflecting the local end-users behaviour

  • Popular end-users, whose content is consumed worldwide, should have a

different way of working than non-popular local end-users, whose content will be locally consumed

  • We perform local routing between the end-users who are in the immediate

vicinity.

  • Locality is defined by network routing hop:

– Two users are local if there are separated by 2 routing hops (or any other value depending on the design configuration).

  • A centralized (SDN-based) controller allows to dynamically configure the NDN

forwarding tables, based on the social interactions in the OSN:

– e.g. add route in the local NDN routers if close friends/followers are on-line, remove it if not.

slide-8
SLIDE 8

8

Publication of a Content

  • For naming end-users and their contents (text messages, videos,

photos, etc), we suggest the following hierarchical naming:

– /Twitter/UserXXX/TweetAAA, /Twitter/UserXXX/VideoBBB

  • We consider local network regions of 2 routing hops.

Twitte ter r server NDN controller ller (SDN (SDN-like like archite tectu cture)

User er Joe e is gett ettin ing g onlin ine

NDN forwardin ding g table (NDN DN FIB) Inform m the SDN controller ller to config igur ure e new routes Add the route to name prefix /Twitter ter/J /Joe

  • e/ into the local

l NDN FIBs NDN FIB User Walter User Bob

/Twit itter/ r/Joe/ Joe/ /Twit itter/ r/Joe/ Joe/

Joe e publis ishes hes a cont nten ent

/Twitter/Joe/Video10

User Alice

slide-9
SLIDE 9

9

Local Routing for Retrieving Content from Local Users

Twitte ter r server NDN controller ller (SDN (SDN-like like archite tectu cture) NDN forwardin ding g table (NDN DN FIB) NDN FIB Local user Alice ce request sts s Joe’s content User Bob

/Twit itter/ r/Joe/ Joe/ /Twit itter/ r/Joe/ Joe/ /Twitter/Joe/Video10

User Walter User Joe is online and can serve his contents

Inter terest est (/Twit witte ter/ r/Jo Joe/ e/Vid ideo eo10 10)

Joe sends content t Data back to Alic ice e

Cache Cache

  • Local users (2 routing hops far away from Joe) will get content

directly from Joe, instead from the Twitter server

slide-10
SLIDE 10

10

Using Twitter Server for Retrieving Content from Non-Local Users

Twitte ter r server NDN controller ller (SDN (SDN-like like archite tectu cture) NDN forwardin ding g table (NDN DN FIB) NDN FIB User Alice

/Twit itter/ r/Joe/ Joe/ /Twit itter/ r/Joe/ Joe/ /Twitter/Joe/Video10

  • Non-local users will be served by the Twitter server

User Walter User Joe is online and can serve his contents

Inter terest est (/Twit witte ter/ r/Jo Joe/ e/Vid ideo eo10 10)

Non Non-loc

  • cal

al user Bob requests Joe’s content

Use e prefix fix /Twitt witter er to rout ute e the e Inter terest est to the e serv rver

slide-11
SLIDE 11

11

Economic Incentives for the local-aware ICN- based content delivery

  • Estimate the economic viability for each involved stakeholder when

adopting our locality-aware and NDN-based routing scheme for delivering content

  • Compare current delivery chain vs local-aware ICN-based delivery

– with involved actors : – Service Providers – Delivery Service Provider provider) – Access ISP

  • Variables :

– the ratio of users (e) which can retrieve content from local users – the probability of cache hit (p)

slide-12
SLIDE 12

12

Economic Incentives for the local-aware ICN- based content delivery

  • Exemple : Cost for Service Provider :

– With local-aware ICN-based delivery COSN = [(1-p].D.(1-e).F .W.T] + [(1-p].D.(1-e).P .T] + [gp].D.F .T] + [tf.T] + [p.D.tu.T] network cost to serve cache miss processing cost for handling cache miss cost for delivering content from caches transaction costs for outsourcing delivery

slide-13
SLIDE 13

13

Economic Incentives for the local-aware ICN- based content delivery

  • Exemple : Cost for Access ISP

– With local-aware ICN-based delivery COSN = [(1-p].D.(1-e).F .B.T] + [(1-p].D.(1-e).F .XE->T.T] + [SE.F .T] + [PE.D.T] + [p.D.KE.T] + d transport cost for cache misses transit cost for cache misses storage cost processing cost Accouting and billing cost additional cost

the ratio of users (e) the probability of cache hit (p)

slide-14
SLIDE 14

14

Conclusion

  • ICN can be deployed if actors can keep some of their business
  • We envision a progressive deployment of ICN, with relationships

between actors to set up

  • For a network point of view (network traffic), a local-aware delivery

is interesting

  • The economic evaluation we performed proved the interest of a

local-aware NDN-based delivery architecture for most of the involved actors.

slide-15
SLIDE 15

Thanks. Q&A