Performance Evaluation of Peer-to-Peer Video Streaming Systems

1
Performance Evaluation of Peer-to-Peer Video
Streaming Systems

• Wilson, W.F. Poon
• The Chinese University of Hong Kong

2
Content

• Introduction
• Related Works
• System Model
• Experimental Results
• Conclusion

3
Introduction (1)

• Providing video streaming services have long been
  a research topic
• parallel server designs such as RAID
• multicast/broadcast transmission schemes
• distributed VoD systems
• Tremendous growth in computer power of personal
  computers
• peer-to-peer (p2p) systems
• Peers contribute storage, content and bandwidth

4
Introduction (2)

• Most of these p2p systems have been developed for
  file sharing/web caching services
• Search mechanism
• Storage management
• Maximize file availability or system reliability
• The work on p2p video streaming has not been
  thoroughly studied
• Investigate whether such a p2p system is
  applicable to supporting video streaming
  applications
• Distributed data storage and its impact on
  streaming performance
• Analytical framework incorporated the effect of
  data replication and placement policies

5
P2P Streaming Systems (1)

• One of major challenges of a p2p system
• Peer machines may be turned on and off in an
  unpredictable manner
• The system experiences very worse availability

6
P2P Streaming Systems (2)

• A network has G peers in which I peers (serving
  peers) stores a set of J different videos
• The other peers (free riders) just make requests
  but not contribute their resources
• Assume
• ? is the up probability of the peers
• Tup is the mean up time duration
• Tdown is the mean down time duration

• Assume
• Ni is the amount of shared storage in peer i
• bj is the size of video j
• qj is the request probability for video j
• Cj is the bit rate for video j

7
P2P Streaming Systems (3)

• nj is the number of replicas for video j, vj
• Requests to a serving peer for vj is given by

8
System Availability

• With full replication scheme
• The video j is not available when all the peers
  storing vj are off-line simultaneously
• System Availability

9
System Arrival (1)
Playback is unsuccessful if the request is blocked
rejected
peer
Availability
peer
requested video available
peer
10
System Arrival (2)

• New requests to peer i

Vi Set of videos stored in peer i
11
System Arrival (3)

• Assume
• Service time (video length) follows an
  exponential distribution
• up duration is exponentially distributed

• Probability of requests redirected by the up
  peer

L mean video length Tup mean up time duration
12
System Arrival (4)

• Total partially served traffic

• Redirect requests to peer k

13
System Arrival (5)

• where

14
System Arrival (6)

• The equations can be represented

• Redirect arrivals can be solved

15
System Blocking (1)

• Unsuccessful playback
• Proportion of requests that cannot completely
  playback the whole video
• Assume
• Poisson Arrival Process
• Video length, up and down durations follow
  exponential distribution
• States of peer i can be represented by a Markov
  Model

16
System Blocking (2)
OFF
ON/0
ON/1
ON/2
ON/K
Peers state diagram
17
System Blocking (3)

• Since a peer will not receive any requests
  (new/redirect) in off state, the probability of
  requests blocked by a peer is equal to

18
System Blocking (4)

• A new video request may be redirected by peers
  several times to finish the video playback
• If either the new request or the redirected
  request is blocked, the playback is considered to
  be unsuccessful

19
Experimental Results

• Simulation is built to verify the model
• Randomly determine the number of replicas for
  each video (random replication)
• Randomly store the replicas among peer (random
  placement)
• Video popularity follows a Zipf distribution with
  parameters 0.271
• Mean video length is 2 hours
• Tup Tdown 10 hours
• Measure the unsuccessful playback rate
• Peers cannot complete the video playback

20
Results Arrival Rate

• Number of peers1200
• Number of videos200
• Video length7200s

21
Results Serving Peers

• Arrival rate0.04/s
• Number of videos200
• Video length7200s

22
Results Peer Availability

• Arrival rate0.02/s
• Number of peers1200
• Number of videos200
• Video length7200s

23
Replication Strategy - MinReq

• For video streaming, a request that can be served
  requires
• The requested video is available in the system
• The serving peers have the available bandwidth
• Determine the number of video replicas by
  minimizing the load of the serving peers

Minimize
Subject to

• Optimal replication profile

24
Results Serving Peers (MinReq)

• Arrival rate0.04/s
• Number of videos200
• Video length7200s
• Peer Storage10

25
Error on Video Popularity

• Considering an estimation error

• Estimated popularity is used to generate the
  replication profile

26
Results Estimation Error

• Arrival rate0.04/s
• Number of videos200
• Video length7200s

27
Conclusion

• Consider the performance of a p2p system for
  video streaming services
• Evaluate data storage and its impact on video
  streaming
• Develop analytical framework to capture the
  properties of the system
• Data replication
• placement policy
• Optimal replication scheme may not significantly
  improve the successful playback rate