Dynamic service deployment @ Newcastle

In the last month or so we have started an effort here in Newcastle at NEReSC to build and deploy a production service for our application scientists. We want to utilise all the computational resources around the campus and eventually integrate our infrastructure with the UK National Grid. Of course that’s the ultimate step. For now we are taking baby steps.

This of course has been done in other places (e.g. White Rose Grid, LeSC, Cambridge, etc. and I am sure in many places in the States and elsewhere) using technologies like N1 Grid Engine, Condor, Globus Toolkit, etc. We want to deploy something similar.

However, at the same time as this effort, we also want to investigate the possibility of treating such an infrastructure as a large application server that can host the implementation of any service logic by dynamically deploying it based on policy-, load-, security-based information. Effectively the idea is to convert a distributed computational facility to a large service logic hosting environment.

Chris Fowler has been doing great stuff as part of his PhD work in this area (with a focus on the security aspects) and leading the design of the overall architecture. Charles Kubicek has been looking at deployment and reconfiguration issues in a cluster environment using Condor which is going to be used underneath our service. Arijit Mukherjee and John Colquhoun have been looking at what’s involved in the implementation side using Axis and they also contribute to the overall architecture.

The team is great and is really fun talking with them to see what they are doing. This short-term, internal effort is overseen by Paul Watson who, as always, is responsible for the overall vision and direction.

The aim is to deliver something by January/February. I’ll blog more about it then.

Apart from being a great and fun effort that I felt it deserved to be mentioned and it also relates to a small investigation I did as a result of the ideas discussed in the few meetings we have had so far. More about this in the next blog entry.