Can Internet VideoonDemand Be Profitable

1
Can Internet Video-on-Demand Be Profitable?

• Cheng Huang, Jin Li (Microsoft Research), Keith
  W. Ross (Polytechnic University)
• ACM SIGCOMM 2007

2
Outlines

• Motivation
• Trace User demand behavior
• Peer assisted VoD
• Theory
• Real-trace-driven simulation
• Cross ISP traffic issue
• Conclusion

3
Motivation

• Saving money for huge content providers such as
  MS, Youtube
• Video quality is just acceptable

User demand
User BW
User BW
User BW
Traffic
ISP Charge
Traffic
ISP Charge
Traffic
ISP Charge
Traffic
ISP Charge
Client Server
P2P
Video quality
Video quality
Video quality
Video quality
4
P2P Architecture

• Peers will assist each other and wont consume
  the server BW
• Each peer have contribution to the whole system
• Throw the ball back to the ISPs
• The traffic does not disappear, it moved to
  somewhere else

5
Outlines

• Motivation
• Trace User demand behavior
• Peer assisted VoD
• Theory
• Real-trace-driven simulation
• Cross ISP traffic issue
• Conclusion

6
Trace Analysis

• Using a trace contains 590M requests and more
  than 59000 videos from Microsoft MSN Video (MMS)
• From April to December, 2006

7
Video Popularity

• The more skewed, the much better

8
Download bandwidth

• Use
• ISP download/upload pricing table
• Downlink distribution
• to generate upload bw distribution

9
Demand V.S. Support
10
User behavior - Churn
11
User Behavior - Interaction
12
Content quality revolution
13
Traffic Evolution
2.27
1.23
Quality Growth 50 User Growth
33 Traffic Growth 78.5
14
Outlines

• Motivation
• Trace User demand behavior
• Peer assisted VoD
• Theory
• Real-trace-driven simulation
• Cross ISP traffic issue
• Conclusion

15
P2P Methodologies

• Users arrive with poison distribution
• Exhaustive search for available upload BW

Video rate 60
60
70
Total Demand 60 x 4 240
100
30
40
0
10
0
Total Support 1004030100 270
40
0
100
16
System status

• If Support gt Demand
• Surplus mode, small server load
• If Support lt Demand
• Deficit mode, VERY large server load
• If Support Demand
• Balanced mode, medium server load

17
Prefetch Policy

• When the system status vibrates between surplus
  and deficit mode
• Let every peer get more video data than demand
  (if possible) in surplus mode
• And thus they can tide over deficit phase

18
Outlines

• Motivation
• Trace User demand behavior
• Peer assisted VoD
• Theory
• Real-trace-driven simulation
• Cross ISP traffic issue
• Conclusion

19
Methodology

• Event-based simulator
• Driven by 9 months of MSN Video trace
• Use greedy prefetch for P2P-VoD
• For each user i, donate its upload BW and
  aggregated BW to user i1
• If user is buffer point is smaller than user
  i1s
• BW allocate to user i1 is no more than user i

20
Trace-driven simulationLevel

• Non-early-departure Trace
• Non-user-interaction Trace
• Full Trace

21
Simulation Non-early-departure
22
Simulation Early departure (No interaction)

• When video length gt 30mins, 80 users dont
  finish the whole video

23
Simulation Full

• How to deal with buffer holes
• As user may skip part of the video
• Two strategies
• Conservative Assume that user BW0 after the
  first interaction
• Optimistic Ignore all interactions

24
Results of full trace simulation (1/2)
25
Results of full trace simulation (2/2)
26
Outlines

• Motivation
• Trace User demand behavior
• Peer assisted VoD
• Theory
• Real-trace-driven simulation
• Cross ISP traffic issue
• Conclusion

27
ISP-unfriendly P2P VoD

• ISPs, based on business relations, will form
  economic entities
• Traffic do not pass through the boundary wont be
  charged
• ISP-unfriendly P2P will cause large amount of
  traffic

28
Simulation results of unfriendly P2P
29
Simulation results of friendly P2P

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

30
Good for the paper

• Large scale on-demand video streaming system
  measurement
• Simulation to show peer assistance can
  dramatically reduce server bandwidth cost
• Pointing out and try to solve impact of
  peer-assisted VoD on the cross-traffic ISPs
• A model to explain simple operation mode of
  peer-assisted VoD
• Comparison of three natural pre-fetching
  policies non pre-fetching, water-level and greedy

31
Bad for the paper

• Too simple conclusion for the user upload
  bandwidth breakdown
• Simple model for peer assisted VoD
• ISP friendly Peer-assisted VoD is most likely
  impossible to study and apply
• Only study peer-assisted VoD based on pure VoD
  System
• Not so many impressive results from measurements

32
What we can

• NAT problem might solve by locality information
• Any other models can explain more factors about
  VoD system or VoD system with peer assist
• User interaction and peer churn in the Grid2.0
  system are two interesting topic to study
• QoS of peer and server cost inside peer assisted
  VoD are some direction for research

33
Thank You