I'm doing research into peer to peer computing, with a special interest in using Distributed Hash Tables to provide reliable self-healing services. The particular services I'm interested in are those associated with grid computing, which is a paradigm for distributed computing that intends to make access to distributed storage and cpu cycles as simple for the end user as using local resources.

The targets of grid computing are ambitious, particularly as it aims to coordinate access to tens of thousands of computers. It is a technology that will play a vital roles in many areas, not least in processing the data from important new physics experiments such as the LHC and one day the ILC.

To this end, I have been looking at how to reliably store information in a hash table, to produce the kind of information service that might scale to serving tens or hundreds of thousands of clients while being resilient to correlated failures. The work I have done so far has mainly been based in simulation, although we have recently produced a prototype implementation of some of the ideas which is used as part of SAM-Grid. Further details in the papers below.

I have also worked on a large scale scalable event based monitoring system, and a framework for running both stress and unit tests on distributed systems. Details of both can be found in the papers below.

Publications
Replication Strategies for Reliable Decentralised Storage
In Proceedings of DAS-P2P06 (PDF)

SAMGrid Peer-to-Peer Information Service
In proceedings of CHEP06 (PDF)

The SAM-Grid monitoring service and its integration with monalisa.
In Proceedings of CHEP04 (PDF)

Application of the SAM-Grid test-harness for performance evaluation and tuning of a distributed cluster implementation of data handling services.
In Proceedings of CHEP04 (PDF)

Publications I am named on, but am not primary author of
Fermilab Computing Divison
From CHEP