A Modeling Framework to Understand the Tussle between ISPs and - - PowerPoint PPT Presentation

A Modeling Framework to Understand the Tussle between ISPs and Peer-to-Peer File Sharing Users IFIP Performance 2007

Michele Garetto – Univ. of T

• rino

Daniel Figueiredo – Fed. Univ. of Rio de Janeiro Rossano Gaeta – Univ. of T

• rino

Matteo Sereno – Univ. of T

• rino

IFIP Performance 2007

An Internet Tale

 Once upon a time...

 user unhappy (“world wide wait”)  ISP unhappy (little revenue)

user

Internet ISP

 Then came broadband access...

user

Internet ISP

 And they lived happily ever after...

fast!

IFIP Performance 2007

The Villain Arrives

 P2P file sharing application

(Kazza, Bittorrent, Emule, etc)

user

 users love it!  good and free

content, overnight downloads

ISP

 ISPs hate it!  users using

their link

 Internet link

utilization gone wild

 degrades all

subscribers

 more

bandwidth costs money!

Internet ISP

IFIP Performance 2007

Taking Care of The Villain

 Mafia

style!

user

 seriously threaten

application developers!

 doesn't seem to work

(Napster story)

Is it Really a Villain?

 Users love it!  Driving force for broadband adoption  Increased revenue for ISPs

IFIP Performance 2007

Some Other Options

user

 User unfriendly ideas

 increase

subscription cost

 volume based

pricing

 block / shape P2P

traffic

 User friendly ideas

 acquire more

bandwidth

 network caching  application-layer

redirection

What should the ISP do?

ISP

???

IFIP Performance 2007

The Real Thing (Data)

P2P represented 60% of Internet Traffic at the end of 2004!

IFIP Performance 2007

Our contribution

 Modeling framework to analyze

interactions between P2P file sharing users (their traffic) and ISP

 economic + performance models

 Basic insights about system

dynamics

 Used to evaluate different

strategies to manage P2P traffic

IFIP Performance 2007

Meet the Players

 generates

queries

 quality of service

expectations

 what's hot,

what's not

user

ISP

 goal: to make

money!

 sets subscription

price

 controls

bandwidth

 influences P2P

• app. behavior

 P2P application  locates object

Network

 network

architecture

 protocols

IFIP Performance 2007

System Setting

User issues query

 Bd  constrained resource for ISP  Outside download consumes Bd

number of P2P users within ISP number of P2P users outside ISP

IFIP Performance 2007

Simple System Model

• bject retrieval probability:
• prob. P2P app.

locates object prob.object is located inside ISP Model for “Internet to ISP” link aggregate query rate unconstrained downloads from within the ISP

IFIP Performance 2007

User Utility Function

 Satisfaction model for user i

• 1
• 0.5

0.5 1 0.2 0.4 0.6 0.8 1σ User utility

probability of successful

• bject retrieval

shape parameter subscription cost Minimum service level for user i user

IFIP Performance 2007

ISP Utility Function

 Profit for ISP (revenue - costs)

fixed charge cost per unit of external bandwidth revenue from subscribers’ fee

 The ISP starts service only if

ISP

IFIP Performance 2007

Modeling Traffic Locality

 Probability there exis

t at least one internal replica

• f object replicated r times in the system

Number of internal copies Number of external copies

 Probability to download from internal replica

locality parameter

IFIP Performance 2007

Analytical Results

 How much bandwidth should the ISP buy

to minimally satisfy the users?

Bmin=max [0,nqmin−q r n/N]



identical users and n >> N

 Non-linear behavior (on n)

 more users, more locality, less BW needed  can be zero if n large enough

 May not yield profit

 too few users, too costly to satisfy them

 Dependent on multiple parameters

IFIP Performance 2007

Impact of Object Replication (r)

1000 2000 3000 4000 5000 10000 20000 30000 40000

Bmin (objects/day) Number of users, n r = 500 r = 1000 r = 1500

 more replicas, better locality, lower Bmin

more bw needed to support larger user population less bw needed (users satisfied locally)

IFIP Performance 2007

Impact of Subscription Cost (c)

• 10000

10000 20000 30000 40000 5000 10000 15000 20000 25000 30000

c = 0.25 c = 1.0 c = 1.4 c = 1.6 Number of users, n Utility of ISP

nmin

 critical mass of users, nmin  lower cost, more profit earlier, less profit later

ISP does not provide service!

IFIP Performance 2007

Critical Mass of Users, nmin

20000 40000 60000 80000 100000 120000 140000 200 400 600 800 1000

nmin

Average object replication, r

β2 = 6 β2 = 5 β2 = 4 β2 = 3

Cost per unit of bandwidth for ISP

 higher bw cost for ISP, higher critical mass  large influence of number of replicas

IFIP Performance 2007

Model Refinements

 Simple model

 users' access

bandwidth are unconstrained

 object replication

is a parameter

 all objects are

identical (no popularity)

 users availability

identical

 Refined model

relax these assumptions

 propose object

popularity and replication model

IFIP Performance 2007

Object Popularity and Replication Model

 Temporal evolution of object popularity  Objects' popularities evolve differently  Objects continuously introduced and

removed by users

 Analytical technique based on Poisson

shot noise process

Number of replicas of an object at time t?

IFIP Performance 2007

Example

• bject

popularity time

A video from the news A popular song

t1 t2

• bject request

 at request time, both have same popularity,

but news has more replicas

IFIP Performance 2007

Limited Bandwidth Refinements

users BW consumption is limited to bd each user within ISP modeled separately users upload bandwidth limited to bu rate of download requests to user i

IFIP Performance 2007

Results from Refined Model

 Degenerate to

simple model

 when parameters

set appropriately

 Other interesting insights

 influence of limited upload bandwidth  upload/download bandwidth asymmetry  object popularity and replication  influence of user impatience

1000 2000 3000 4000 5000 6000 7000

5000 15000 25000

bd = 2000

Bmin

Number of users, n

bd = 1000 bd = 500 bd = 100 bd = 10

IFIP Performance 2007

Impact of asymmetric access bandwidths

2000 4000 6000 8000 10000 12000 500 1000 1500 2000 2500 3000 3500 4000

Upload bandwidth, bu (for fixed number of users = 20000) bd = bu bd = 2 bu bd = 3 bu bd = 4 bu

Bmin (objects/day)

 cost for ISP increases as ratio increases  better if upload BW is greater than download

IFIP Performance 2007

Conclusions

 Development of simple analytical model

 economics + performance  interaction between P2P users (their traffic)

and ISP

 insights into strategy for ISP to manage its

traffic

 Model for object popularity and

replication

 of independent interest

 Future work

 Multiple ISPs competing with each other

IFIP Performance 2007

THE END

 Thank you!  Questions? Comments?

CacheLog ic

Advanced Solutions for P2P Networks

Presentation | WCW 2005

Impact on Service Providers

• P2P is the

the dominant protocol

• In excess of 92% of P2P traffic crosses transit/peering links
• P2P protocols will aggressively consume

aggressively consume any available bandwidth capacity

• Due to P2P’s symmetrical nature on average 80% of upstream

capacity is consumed by P2P

• P2P affects QoS levels for ALL subscribers
• Service Providers can not afford to block or restrict P2P
• ISPs must intelligently manage P2P - blocking and shaping

doesn’t work

P2P is driving consumer broadband uptake P2P is driving consumer broadband uptake … …and broadband is driving P2P uptake and broadband is driving P2P uptake

HT T P O ther Non P2P O ther P2P FastT rack eDonkey BitT

• rrent

Recognising Gnutella

IFIP Performance 2007

The ISP perspective vs P2P:

threat or opportunity ?

 P2P traffic: friend or foe ?

 friend: driving force for adoption of

broadband access by the users

 foe: overwhelming amount of traffic

 What is the best strategy to manage

P2P traffic in my network ?

 Try to kill it ?  Do nothing ?  Educate it ? How ?

IFIP Performance 2007

Strategies to manage P2P traffic

 Acquire more bandwidth  Block P2P traffic  Traffic shaping (e.g., priority to non-P2P)  Pricing schemes based on user traffic

volumes / bandwidth caps

 Network caching / customized P2P

application within ISP

 Application-layer redirection of P2P

traffic

IFIP Performance 2007

 Assumption: n identical users  N = 50 millions  request rate by user (object/day)  introduction of new contents by

user (object/day)

Results

Minimum required external bandwidth:

IFIP Performance 2007

The impact of efficacy in exploiting traffic locality ( )

20000 40000 60000 80000 100000 0 100000 300000 500000 700000

= 0.25 = 0.5 = 0.75 = 1.0

Number of users, n Minimum required bandwidt h Bmin

(objects/day)

IFIP Performance 2007

Model refinements

 Impact of finite bandwidth of the users

 In case of constant traffic load, the cost for

the ISP increases if it provides more download bandwidth to the users (!)

 The system performance is strongly affected

by the upload bandwidth of the users, that should be larger than or equal to the download bandwidth (contrary to common practices, e.g., ADSL lines !)

The impact of bd

1000 2000 3000 4000 5000 6000 7000 5000 10000 15000 20000 25000 30000

bd = 2000

Bmin

Number of users, n

bd = 1000 bd = 500 bd = 100 bd = 10