P2P Models

P2P Models

1. Centralised Directory Model

   • Peers connect to the central directory and inform the contents they offer for sharing
   • Requires structure for directory Server - Scalability Issues
   • Napster used it
     • Not scalable

2. Flooded Request Model

   • Pure P2P: No central servers No advertisements of resources
   • A peers request is flooded
     • broadcast to its directly connected peers. A request is flooded until it's answered or a maximum number of flooding steps is reached
   • Advantages
     • No central point of failure Limited per-node state
   • Drawbacks
     • Requires a lot of bandwith
   • Used by Gnutella
     • Not efficient Search

3. Document Routing Model

   • Most recent approach Also pure p2p - no central servers
   • Each peer that joins the system is assigned a random ID Number
   • Each peer knows the number of peers
   • Publication:
     • When resources are published an ID is assigned to it based on a hash of the document's contents and name.
     • 2. Each peer will route the document towards the peer with the most similar ID
     • 3. Process is repeated until nearest peer ID is the current peer's ID  
     • Each routing operation ensures that a copy of the document is kept
   • Request:
     • 1. When a peer requests a document from the system, the request goes to the peer with most similar ID
     • 2. The process is repeated until a copy of the document is found
     • 3. The document is transferred back to the requested peer
     • 4. Every peer participating in the routing will keep a local copy of the document.
   • Very efficient for large, global communities. However, document ID must be known BEFORE posting a request also might lead to islanding.

4. Issues:

   • Anonymity
     • People should have access to resources without fears (legal, social). Censorship of digital content should not be possible.
     • Three kinds of anonymity: Sender anonymity Receiver anonymity Mutual anonymity

5. Topologies

   • Core
     • Centralised
       • Client/server Web servers Databases Napster search Instant Messaging
     • Ring
       • Fail-over clusters Simple load balancing
     • Hierarchical
       • Tree structure DNS Usenet
     • Decentralised
       • Gnutella Freenet Internet Routing
   • Hybrid
     • Centralised-Ring Centralised-Centralised Centralised-Decentralised