Peer-to-Peer Storage Systems Home Page

Department of
Computer Science
and Engineering

Peer-to-Peer Storage Systems

This is a cached copy. Go to http://www.cse.ogi.edu/~smurthy/p2ps/index.html for the current version.

Related links (external)
- CSE581: Internet Technologies
- Pastry
- OceanStore
- Freenet

Overview

This web is a result of research carried out as a requirement to complete CSE581: Internet Technologies (Winter 2002).

Peer-to-Peer (P2P) storage systems support persistence in a network of peer computers. The underlying network may be defined at any layer, although most recent ones are defined as Internet overlay networks. Computers participating in such a network play one or more of these roles: server, router, client, cache. Servers store data; routers route messages; clients access servers through routers either to store data or retrieve data; caches simply cache data. P2P storage systems must be fast, fault-tolerant, scalable, and reliable. They must possess good locality characteristics or behave like a small world network: keep the clique, and reach everything fast!

The responsibilities/needs of P2P storage systems might be classified into two distinct functions- networking (including routing and locating) and storage management (including caching). Some systems such as Freenet define these functions together, whereas some systems such as PAST and OceanStore define these functions separately. PAST and OceanStore use Pastry and Tapestry respectively for networking. This research studies these four systems.

Summary and discussion

A research summary of the systems studied was presented to the class (MS PowerPoint presentation file). The following issues came up during the presentation. The papers studied answer these questions to varying degrees (mostly not satisfactorily).

Is it possible to lookup data (files/objects) stored using some of its attribute (for example, file name)? Clients may not always know the ID.
What is the performance of the overlay network?
How is consistency of the cached copies maintained?
What is the effort to cache?
What is the cost of identifying and integrating a new node?
What security measures are in place?
How is data updated? What is the cost? What about the consistency of updates?

References

References are listed using the Name-Year system based on Scientific Style and Format: The CBE Manual for Authors, Editors, and Publishers. 1994; 6th edition; Cambridge University Press; New York, NY. Sample usage and other helpful information available online: <http://www.wisc.edu/writing/Handbook/DocCBENameYear.html> Accessed 2002 Jan. 27.