Can Internet Video-on-Demand Be Profitable? Cheng Huang, Jin Li (Microsoft Research), Keith W. Ross (Polytechnic University) ACM SIGCOMM 2007
Outlines n Motivation n Trace – User demand & behavior n Peer assisted VoD –Theory –Real-trace-driven simulation n Cross ISP traffic issue n Conclusion
Motivation n Saving money for huge content providers such as MS, Youtube n Video 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 Architecture n Peers will assist each other and won’t consume the server BW n Each peer have contribution to the whole system n Throw the ball back to the ISPs –The traffic does not disappear, it moved to somewhere else
Outlines n Motivation n Trace – User demand & behavior n Peer assisted VoD –Theory –Real-trace-driven simulation n Cross ISP traffic issue n Conclusion
Trace Analysis n Using a trace contains 590M requests and more than videos from Microsoft MSN Video (MMS) n From April to December, 2006
Video Popularity n The more skewed, the much better
Download bandwidth n Use –ISP download/upload pricing table –Downlink distribution to generate upload bw distribution to generate upload bw distribution
Demand V.S. Support
User behavior - Churn
User Behavior - Interaction
Content quality revolution
Traffic Evolution Quality Growth: 50% User Growth: 33% Traffic Growth: 78.5%
Outlines n Motivation n Trace – User demand & behavior n Peer assisted VoD –Theory –Real-trace-driven simulation n Cross ISP traffic issue n Conclusion
P2P Methodologies n Users arrive with poison distribution n Exhaustive search for available upload BW 100 Video rate: Total Demand 60 x 4 = 240 Total Support = 270
System status n If Support > Demand –Surplus mode, small server load n If Support < Demand –Deficit mode, VERY large server load n If Support ≈ Demand –Balanced mode, medium server load
Prefetch Policy n When the system status vibrates between surplus and deficit mode n Let every peer get more video data than demand (if possible) in surplus mode –And thus they can tide over deficit phase
Outlines n Motivation n Trace – User demand & behavior n Peer assisted VoD –Theory –Real-trace-driven simulation n Cross ISP traffic issue n Conclusion
Methodology n Event-based simulator n Driven by 9 months of MSN Video trace n Use greedy prefetch for P2P-VoD –For each user i, donate it’s upload BW and aggregated BW to user i+1 –If user i’s buffer point is smaller than user i+1’s n BW allocate to user i+1 is no more than user i
Trace-driven simulation Level n Non-early-departure Trace n Non-user-interaction Trace n Full Trace
Simulation: Non-early- departure
Simulation: Early departure (No interaction) n When video length > 30mins, 80%+ users don’t finish the whole video
Simulation: Full n How to deal with buffer holes –As user may skip part of the video n Two strategies –Conservative: Assume that user BW=0 after the first interaction –Optimistic: Ignore all interactions
Results of full trace simulation (1/2)
Results of full trace simulation (2/2)
Outlines n Motivation n Trace – User demand & behavior n Peer assisted VoD –Theory –Real-trace-driven simulation n Cross ISP traffic issue n Conclusion
ISP-unfriendly P2P VoD n ISPs, based on business relations, will form economic entities –Traffic do not pass through the boundary won’t be charged n ISP-unfriendly P2P will cause large amount of traffic
Simulation results of unfriendly P2P
Simulation results of friendly P2P n Peers lies in different economic entities do not assist each other
Conclusion (Pros) n This paper gives a representative trace analysis that breaks the myth of upload BW problems n Successfully address the importance of the P2P cross-ISP problem
Conclusions (Cons) n Weak and unrealistic P2P models n Unclear comparisons between each P2P strategies and simulations
Thank You