P2P-Tuple: Towards a Robust Volunteer Computing Platform

Review

Problem and Proposal

We propose to build the next generation Volunteer Computing systems on top of well studied Peer-to-Peer techniques to fully take advantage of its decentralized characteristic and its very large, shared data storage capacity.

Usually a central server is usually used to assign these jobs to voluntarily contributed machines, synonymously known as volunteers, workers or executors. So the overall availability of the system depends on the master servers.

The main contribution of the paper is the proposed architecture of P2P based VC platform called P2P-Tuple (which removes all central components).

Related Works

  • CCOF - P2P based VC platform in which member peers construct a CAN DHT overlay. Results from volunteers would be returned to Application Scheduler for validation and storage (AS has the same properties as BOINC’s central server)
  • TONIC - pioneer in Tuple Space’s application in parallel processing. Central server is used to host Tuple Space for storing jobs and results, Lookup service is also provided on that server to allow discovery and accessing of the Tuple Space.
  • P2P-Tuple’s Architecture

Picture above could easily explain the difference between 1st, 2nd and 3rd system.

Design of P2P-Tuple

This system is completely decentralized. In P2P-Tuple, a single Pastry DHT overlay is formed (each volunteer node becomes a Pastry peer). Peers can choose to join one of the SCRIBE based groups (scalable group communication system). At the same time P2P Object Storage System uses PAST to form the Tuple Space. Data stored in this service is associated with a key and expiration date, large files are divided into chunks and encoded using IDA algorithm.

Proved that developed system based on usage of mirrors as well works better existing ones.

Open source realization is also available.

decentralized_storage_systems/p2ptuple.txt · Last modified: 2012/04/23 01:07 by julia
 
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 3.0 Unported
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki