CongestionDistortion Optimized PeertoPeer Video Streaming

1
Congestion-Distortion Optimized Peer-to-Peer
Video Streaming

• Eric Setton, Jeonghun Noh and Bernd Girod
• Information Systems Laboratory
• Stanford University
• Recently joined HP Labs

2
Outline

• Motivation
• Optimized scheduling for peer-to-peer (CoDiO P2P)
• Prioritization algorithm sender ? receivers
• Retransmission scheduler receiver ? senders
• Experimental Results
• Comparison to state-of-the-art schedulers
• Benefits of adaptive scheduling on P2P streaming

3
Prior Work

• TCP-friendly rate-control
• Indicates average rate as a function of collected
  statistics
• Does not indicate any particular schedule
• Rate-Distortion optimized scheduling (RaDiO)
• Formalization of the multimedia scheduling
  problem
• Adapted the framework to video streaming
• Congestion-Distortion optimized scheduling
  (CoDiO)
• Congestion defined as end-to-end delay
• Designed for throughput-limited streaming
• Same R-D performance as RaDiO and 40 less
  congestion

Floyd et al., 1997
Chou and Miao, 2001
Kalman and Girod, 2003
Setton and Girod, 2004-2006
4
Video Multicast over P2P Networks

• Challenges
• Limited bandwidth
• Delay due to multi-hop transmission
• Unreliability of peers
• Related work
• Chu, Rao, Zhang, 2000
• Padmanabhan, Wang and Chou, 2003
• Guo, Suh, Kurose, Towsley, 2003
• Cui, Li, Nahrstedt, 2004
• Do, Hua, Tantaoui, 2004
• Hefeeda, Bhargava, Yau, 2004
• Zhang, Liu, Li and Yum, 2005
• Zhou, Liu, 2005
• Chi, Zhang, 2006
• Our Approach
• Build and maintain complementary multicast trees
• Adapt media scheduling to network conditions and
  to content

5
Specificities of P2P Multicast
6
Principles of CoDiO P2P

• Scheduler which combines
• optimized prioritization algorithm

Sending peer
Receiving peer
Receiving peer
Receiving peer
Decide which packets to send, when and to which
peer, to maximize performance while limiting
incurred congestion
7
Principles of CoDiO P2P

• Scheduler which combines
• optimized prioritization algorithm
• optimized retransmission requests

Sending peer
Sending peer
Sending peer
Receiving peer
Decide which missing packets to request, when and
from which peer, to maximize performance while
limiting incurred congestion
8
CoDiO Prioritization

• Scheduler iteratively selects
• Intervals between transmission sufficient to
• Mitigate any congestion of the uplink
• Reserve rate for control traffic

P
I
B
P
P
B
B
7 1 6 1 4 1 2
Sender
9
Distortion-OptimizedRetransmission Requests
Parent on multicast tree 1

• Determine missing packets
• Iteratively request most important packet
• Limit number of unacknowledged retransmissions

10
Server-Client Scheduler Performance
Foreman _at_ 290 kb/s
Salesman _at_ 365 kb/s
Simulations over ns-2 2-hop network path, ACKs
from receiver Throughput 400 kb/s, delay 50 ms,
packet losses 2 H.264 encoding
11
Experimental Setup

• Network/protocol simulation in ns-2
• 300 active peers
• Random peer arrival/departure average ON (4.5
  min) / OFF (30 sec)
• Typical access bandwidth distribution
• Over-provisioned backbone
• Delay 5 ms/link congestion
• Video streaming
• H.264/AVC encoder
• 15 minute live multicast
• CIF resolution
• 16-frame GOP I-B-B-B-P

12
Benefits of Optimized Scheduling (I)
Foreman
Salesman
1 dB
2 dB
Simulations over ns-2, 300 peers Number of
trees 4 Retransmissions enabled
13
Benefits of Optimized Scheduling (II)
Foreman _at_ 290 kb/s
Salesman _at_ 320 kb/s
Simulations over ns-2, 300 peers Number of
trees 4 Retransmissions enabled
14
Average Video Sequence for 64 Peers
CoDiO P2P 33.71 dB
No prioritization 30.17 dB
0.8 second latency for all peers
15
P2P Video Multicast 36 of the Peers
CoDiO P2P 33.71 dB
No prioritization 30.17 dB
0.8 second latency for all streams
16
Conclusions

• Summary
• Congestion-distortion optimized scheduling for
  P2P
• Prioritization adapts to the content and to the
  network conditions
• Distortion-optimized retransmission scheduler
• Largest gains when streaming
• with short playout deadlines (up to 4-5 dB)
• close to throughput limit (up to 1-2 dB)
• Future work
• Study P2P scheduling of scalable video
• Release of a peer-to-peer streaming client