Substreaming Trading: Towards an Open P2P Live Streaming System

1
Substreaming Trading Towards an Open P2P Live
Streaming System

• Zhangye Liu, Yanming Shen, Keith W. Ross,
  Shivendra S. Panwar, Yao Wang
• Appear in IEEE ICNP 2008
• Presented by Yan Ding

2
Background and Motivation

• What is Open?
• Source code of the baseline implementation is
  available
• Enable developers to create independent client,
  seed and tracker implementation
• Allow user-generated content
• What is P2P?
• Peers are equal, both Clients and Servers
• Diverse connectivity between peers (Based on the
  content availability info etc.)
• Data Transferred through these connected links
• Use cumulative bandwidth of peers, rather than a
  central sever resource

3
Background and Motivation(contd)

• What is the main issue in Open P2P System?
• Free-rider peers download content but upload at
  low rate, or even without uploading
• How to deal with it?
• Need an incentive mechanism encourage peers to
  contribute their upload bandwidth

4
Main Idea

• BitTorrent
• Open, Peer-Assisted File Distribution
• Incentive principle (tit-for-tat)
• A peer contributes more upload bandwidth,
  get file faster
• Substream Trading
• Open, Peer-Assisted Live-Video Streaming
• Incentive principle (tit-for-tat)
• A peer contributes more upload bandwidth,
  see higher quality video

5
System Design

• Tree or Mesh?
• Tree
• Video Source providers locate at the root, and
  pump video chunk through trees
• Tight relationship between peers (Parents and
  Children)
• Efficient, but not robust to peer churn and
  dynamics
• Mesh
• Peers self-organized based on their available
  content
• Video transferred through partners (a link
  between them)
• robust to peer churn and dynamics
• Used in all large-scale deployment (PPLive,
  PPStream etc.)

6
System Design(contd)

• Chunk or Substream?
• Chunk
• Video is divided into chunks, each contains 1-4s
  video
• Has a playback lag and overhead trade-off
• Substream
• Video is divided into substreams(contain several
  chunks) using Time Dividing
• Reduce the playback lag and overhead

1
2
3
3
3
2
2
1
1
Chunk Streams
1 2 3 4 5 6 7 8 9
7
Mesh-Substreams System

• How does it work?
• Peer discovery Determine existance of peers and
  which substreams they have
• Partner Selection Determine who are the partners
  and establish the link
• Substream Selection Determine which substreams
  should be obtained from which partners

8
Partner Selection

• Select scheme
• Exchange substream map, get Sp,Sq
• Agreement P choose Q and Q choose P
• Establish link
• Monitor scheme
• Monitor service quality
• use Leaky Bucket
• Adapt bad partners

Sp
Sq
SpgtSq For Q Agree For P p(S-Sp)/S
P
Q
9
Substream Selection
Check substream availability
Abstract substream map
Formulate as Optimization Problem
Check chunk availability
Max s.t
Exchange one substeam, if P exchange k substreams
with Pn, consider Pn as to be k virtual ones Pn1,
Pn2..Pnk
Wsweight of each substream. Depends on video
coding scheme (single-layer, layered (Ws2S-s),
MDC)
10
Simulation Setting

• Peer Upload Bandwidth Distribution (Table I)
• Choose 5 types to illustrate
• 256(lowest), 3000(highest), 448(mediate),
  640(fewest), 1024(most)
• Based on two systems
• Single-layer and Layered sys.

11
Simulation Single-Layer Sys.
Free-Riders
Underloaded
Overloaded
Scenario1 Exist altruistic peers Scenario2
Provide different video qualites

12
Simulation Layered Sys.
Differentiated services More upload bandwidth,
more layers received, higher received video
rate.
13
Conclusion

• First Open P2P live system that address incentive
  issue.
• The main idea is inspired by BT, very simple and
  straightforward.
• First P2P live streaming framework supports
  different video coding schemes.
• Design of Mesh and Substream is not a novel
  idea.