A Framework for Cost-Effective Peer-to-Peer Content Distribution

1

• A Framework for Cost-Effective Peer-to-Peer
  Content Distribution
• Mohamed Hefeeda and Bharat Bhargava
• Department of Computer Sciences
• Purdue University
• In ACM Multimedia 2003 Doctoral Symposium
• Thursday November 6th, 100500PM

2
Motivation

• Lots of underutilized end systems (peers)
  connected to the Internet
• Success of Peer-to-Peer (P2P) paradigm
• Kazaa, Gnutella, SETI_at_HOME,
• Fairly high cost for distributing digital
  contents (large videos)
• Why not share? Everybody benefits!
• We propose
• A Collaborative P2P Framework for Content
  Distribution
• Peer contributes little, but we have too many of
  them!

3
Motivation (contd)

• Two settings for the framework
• Infrastructure
• Content provider employs peers resources to
  disseminate contents
• Cooperative resource sharing environment
• Peers cooperate/coordinate to serve requests
• What we gain
• Cost-effectiveness (for supplier client)
• Ease of deployment (on end systems)
• Availability (large degree of redundancy)
• Scalability (more peers ? more resources)

4
Motivation (contd)

• Research problems
• Select and match multiple supplying peers with a
  requesting peer
• Aggregate and coordinate contributions from peers
• Adapt to peer failures and network conditions
• Disseminate contents into the system
• Consider peers rationality (self-interest) into
  protocols
• Assess and incorporate peers trustworthiness

5
Our Contributions

• Cooperative environment ACM MM03
• CollectCast a middleware layer between P2P
  lookup substrate and applications
• PROMISE a sample application atop of CollectCast
  
• Evaluation simulation and implementation
  (PlanetLab)
• Infrastructure J. Computer Networks, To appear
  03
• Hybrid (super peer) architecture
• Two-level peer clustering
• New Searching and dispersion algorithms
• Evaluation simulation
• Current/Future work
• Consider peer rationality in protocols Tech
  Reports 03
• Consider peer trust/reputation in protocols

6
Cooperative Environment

• CollectCast does
• Infer and label topology
• Select best sending peers for each session
• Coordinate contributions from peers
• Adapt to peer failures and network conditions
• PROMISE is
• P2P media streaming system on top of CollectCast
  to show its merits

7
P2P Infrastructure

• Target environments
• Streaming movies to many clients
• Distance learning, media center, corporate
  streaming,
• Key ideas/techniques
• Peer Organization
• Two-level peer clustering
• Manage Join, leave, failure, and super peer
  selection
• Cluster-Based Searching
• Find nearby suppliers, within the cluster
• Cluster-Based Dispersion
• Efficiently disseminate new media files

8
Interested to Know More

• Attend the Doctoral Symposium on Thursday from
  100500PM
• More information and papers at
• http//www.cs.purdue.edu/homes/mhefeeda