can internet video-on-demand be profitable? cheng huang, jin li (microsoft research), keith w. ross...

Post on 17-Jan-2016

216 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Can Internet Video-on-Demand Can Internet Video-on-Demand Be Profitable? Be Profitable?

Cheng Huang, Jin Li (Microsoft Cheng Huang, Jin Li (Microsoft Research), Keith W. Ross Research), Keith W. Ross (Polytechnic University)(Polytechnic University)

ACM SIGCOMM 2007 ACM SIGCOMM 2007

OutlinesOutlines

MotivationMotivation Trace – User demand & behaviorTrace – User demand & behavior Peer assisted VoDPeer assisted VoD

– TheoryTheory– Real-trace-driven simulationReal-trace-driven simulation

Cross ISP traffic issueCross ISP traffic issue ConclusionConclusion

MotivationMotivation

Saving money for huge content Saving money for huge content providers such as MS, Youtubeproviders such as MS, Youtube

Video quality is just acceptableVideo quality is just acceptable

User demand +++

Video quality+++

Traffic+

ISP Charge+Client Server

User BW +

Video quality+

User BW +++

Video quality+++

Traffic++++++++

ISP Charge+++++++P2P

Traffic++

ISP Charge++

User BW ++++++

Video quality+++++++

Traffic+++

ISP Charge+++

P2P ArchitectureP2P Architecture

Peers will assist each other and Peers will assist each other and won’t consume the server BWwon’t consume the server BW

Each peer have contribution to the Each peer have contribution to the whole systemwhole system

Throw the ball back to the ISPsThrow the ball back to the ISPs– The traffic does not disappear, it The traffic does not disappear, it

moved to somewhere elsemoved to somewhere else

OutlinesOutlines

MotivationMotivation Trace – User demand & behaviorTrace – User demand & behavior Peer assisted VoDPeer assisted VoD

– TheoryTheory– Real-trace-driven simulationReal-trace-driven simulation

Cross ISP traffic issueCross ISP traffic issue ConclusionConclusion

Trace AnalysisTrace Analysis

Using a trace contains 590M Using a trace contains 590M requests and more than 59000 requests and more than 59000 videos from Microsoft MSN Video videos from Microsoft MSN Video (MMS)(MMS)

From April to December, 2006From April to December, 2006

Video PopularityVideo Popularity

The more skewed, the much betterThe more skewed, the much better

Download bandwidthDownload bandwidth

Use Use – ISP download/upload pricing table ISP download/upload pricing table – Downlink distribution Downlink distribution

to generate upload bw distributionto generate upload bw distribution

Demand V.S. SupportDemand V.S. Support

User behavior - ChurnUser behavior - Churn

User Behavior - User Behavior - InteractionInteraction

Content quality Content quality revolutionrevolution

Traffic EvolutionTraffic Evolution

2.271.23

Quality Growth: 50%User Growth: 33%Traffic Growth: 78.5%

OutlinesOutlines

MotivationMotivation Trace – User demand & behaviorTrace – User demand & behavior Peer assisted VoDPeer assisted VoD

– TheoryTheory– Real-trace-driven simulationReal-trace-driven simulation

Cross ISP traffic issueCross ISP traffic issue ConclusionConclusion

P2P MethodologiesP2P Methodologies

Users arrive with poison Users arrive with poison distributiondistribution

Exhaustive search for available Exhaustive search for available upload BWupload BW

100

Video rate: 6060

3040

40

0 10

100

0

0

70 Total Demand60 x 4 = 240

Total Support100+40+30+100 = 270

System statusSystem status

IfIf Support Support >> DemandDemand– Surplus mode, Surplus mode, smallsmall server load server load

IfIf SupportSupport << DemandDemand

– Deficit mode, Deficit mode, VERY largeVERY large server server loadload

IfIf SupportSupport ≈≈ DemandDemand– Balanced mode, medium server loadBalanced mode, medium server load

Prefetch PolicyPrefetch Policy

When the system status vibrates When the system status vibrates between surplus and deficit modebetween surplus and deficit mode

Let every peer get more video data Let every peer get more video data than demand (if possible) in than demand (if possible) in surplus modesurplus mode– And thus they can tide over deficit And thus they can tide over deficit

phasephase

OutlinesOutlines

MotivationMotivation Trace – User demand & behaviorTrace – User demand & behavior Peer assisted VoDPeer assisted VoD

– TheoryTheory– Real-trace-driven simulationReal-trace-driven simulation

Cross ISP traffic issueCross ISP traffic issue ConclusionConclusion

MethodologyMethodology

Event-based simulatorEvent-based simulator Driven by 9 months of MSN Video Driven by 9 months of MSN Video

tracetrace Use greedy prefetch for P2P-VoDUse greedy prefetch for P2P-VoD

– For each user i, donate it’s upload BW For each user i, donate it’s upload BW and aggregated BW to user i+1and aggregated BW to user i+1

– If user i’s buffer point is smaller than If user i’s buffer point is smaller than user i+1’suser i+1’s

BW allocate to user i+1 is no more than user BW allocate to user i+1 is no more than user ii

Trace-driven simulationTrace-driven simulationLevelLevel

Non-early-departure TraceNon-early-departure Trace Non-user-interaction TraceNon-user-interaction Trace Full TraceFull Trace

Simulation: Non-early-Simulation: Non-early-departuredeparture

Simulation: Early departure Simulation: Early departure (No interaction)(No interaction)

When video length > 30mins, 80%When video length > 30mins, 80%+ users don’t finish the whole + users don’t finish the whole videovideo

Simulation: Full Simulation: Full

How to deal with buffer holesHow to deal with buffer holes– As user may skip part of the videoAs user may skip part of the video

Two strategiesTwo strategies– Conservative: Assume that user Conservative: Assume that user

BW=0 after the first interactionBW=0 after the first interaction– Optimistic: Ignore all interactionsOptimistic: Ignore all interactions

Results of full trace Results of full trace simulation (1/2)simulation (1/2)

Results of full trace Results of full trace simulation (2/2) simulation (2/2)

OutlinesOutlines

MotivationMotivation Trace – User demand & behaviorTrace – User demand & behavior Peer assisted VoDPeer assisted VoD

– TheoryTheory– Real-trace-driven simulationReal-trace-driven simulation

Cross ISP traffic issueCross ISP traffic issue ConclusionConclusion

ISP-unfriendly P2P VoDISP-unfriendly P2P VoD

ISPs, based on business relations, ISPs, based on business relations, will form economic entitieswill form economic entities– Traffic do not pass through the Traffic do not pass through the

boundary won’t be chargedboundary won’t be charged

ISP-unfriendly P2P will cause large ISP-unfriendly P2P will cause large amount of trafficamount of traffic

Simulation results of Simulation results of unfriendly P2Punfriendly P2P

Simulation results of Simulation results of friendlyfriendly P2P P2P

Peers lies in different economic Peers lies in different economic entities do not assist each otherentities do not assist each other

Conclusion (Pros)Conclusion (Pros)

This paper gives a representative This paper gives a representative trace analysis that breaks the trace analysis that breaks the myth of upload BW problemsmyth of upload BW problems

Successfully address the Successfully address the importance of the P2P cross-ISP importance of the P2P cross-ISP problemproblem

Conclusions (Cons)Conclusions (Cons)

Weak and unrealistic P2P modelsWeak and unrealistic P2P models Unclear comparisons between Unclear comparisons between

each P2P strategies and each P2P strategies and simulationssimulations

Thank YouThank You

top related