edikt Technical Workshop
Date: Friday June 20th 2008 Start: 13:30 Finish: 17:00 Venue: Room 3317 JCMB, Kings Buildings
Purpose of workshop
This edikt workshop will feature a series of talks about research computing around the University. The purpose of the workshop is for researchers to exchange knowledge and experience in research computing amongst interested parties.
The edikt (eScience Data, Information and Knowledge Transformation) project has been running since May 2002 and is using computational science to extract knowledge from vast datasets and simulation models. edikt is funded by the Scottish Funding Council.
For more information on edikt, see the project web site at http://www.edikt.org.
Agenda
1330 Welcome
Mr Terry Sloan
1340 Grid Technologies for SAGES?
Dr Mike Mineter
1410 SPRINT - a Parallel R framework
Dr Jon Hill
1445 COFFEE
1530 eDIKT Brain Imaging Project: Half Time Summary
Dr Paul Armitage
1600 Oncology-II - Genes and Supercomputers
Mr Florian Scharinger
1630 Discussions and Wrap-up
Mr Terry Sloan
1700 CLOSE
Abstracts
Grid Technologies for SAGES? - Dr Mike Mineter
The first part of this talk outlines the goals of SAGES, the Scottish Alliance for Geoscience, Environment and Society, http://www.sages.ac.uk, and how these goals lead SAGES towards use of the ECDF.
In the mid-term SAGES will require additional services, in particular supporting the gathering and use of metadata associated with simulations and their results. The second part of this talk seeks to outline some possibilities for metadata services, and to stimulate discussion concerning the scope for generic metadata solutions to be deployed.
SPRINT - a Parallel R framework - Dr Jon Hill
R is a commonly used tool for many statistical analyses. However, research groups such as the Division of Pathway Medicine are starting to encounter problems with the size of the analyses that need to be carried out, both in terms of size and length of time. HPC is a possible solution to this. Rather than create bespoke applications we can add parallel functionality to R via the SPRINT framework. This has the benefit of requiring very little alteration to existing R scripts, hiding most of the complexity of parallel processing from the researcher whilst gaining the maximum performance by stepping out with R. A test case shows a 6 fold speed-up for a relatively small datasize on 8 processors when carrying out a parallelised correlation analysis.
eDIKT Brain Imaging Project: Half Time Summary - Dr Paul Armitage
This talk summarises the Brain Imaging activities during the first half of the eDIKT project and set out some of the goals for the remaining two years
Oncology-II - Genes and Supercomputers
The Oncology-II project is a collaboration between the University of Edinburgh (represented by EPCC) and the Colon Cancer Genetics Group of the MRC Human Genetics Unit at the Western General Hospital. This project follows on from a previous collaboration investigating the relationship between genetic markers and colorectal cancer.
The aim of this second project is to perform a new analysis consisting of a two-way interaction study for the genetic markers. The data consists of 560,000 markers to be compared against each other, that is, there are of the order of 150 Billion calculations. This requires a significant amount of computational and memory resources.
This talk describes the challenges faced when dealing with such large amounts of data, how nowadays supercomputer enable researchers to perform analysis that have been -- time-wise -- not feasible before and also raises the increasing issue of how researchers can deal with growing amount of result data.
Registration
Attendance at the workshop is free with no prior registration required however the organisers would appreciate it if you could contact Jon Hill (J.Hill@epcc.ed.ac.uk, 0131 651 3396) beforehand with your name.
Useful information
SAGES: http://www.sages.ac.uk/
SPRINT: http://www.edikt.org/edikt2/ParallelRActivity
Brain Imaging: http://www.sbirc.ed.ac.uk/research/research.asp
Oncology: http://www.edikt.org/edikt2/OncologyActivity
For more information please contact Jon Hill: J.Hill@epcc.ed.ac.uk 0131 650 6494