True grid computing may be a resource utilization nirvana, but a group of Melbourne researchers is aiming to complete the final piece of the puzzle via brain research.

Melbourne University's Howard Florey Institute has started using a network of computers to conduct neuroinformatics - an application of computer science, dubbed NeuroGrid, to understand and map the brain - with the intention of enabling worldwide collaboration with other researchers over the Web.

Research leader Dr Gary Egan said the concept of a grid is connecting clusters.

"Our system is a cluster but our software layers are designed to be part of a grid. A lot of development work is needed as the software is not mature but it's beyond a first-look," he said.

"A grid computer is not here yet but we are aiming to achieve this in 12 months; our focus is on the user application end, but we want to close the loop so when things get processed the results are incorporated into the database."

The project began 18 months ago initially with a database of images.

The grid consists of 12 G5 PowerMac and 12 Pentium4 Linux workstations with 3TB of Xserve RAID for storage. Sun workstations were previously used but with the introduction of the Unix-based OS X, the group opted for the Apple systems.

When the desktops are sitting idle they are used for processing - hence forming a grid.

Controlling the grid software involves use of the Sun Grid Engine, Globus middleware to enable remote querying of resources, and a Java workflow engine available to all users to select data sources.

"That's the area that needs customization," Egan said. "We tried to adopt open standards where possible to give the maximum chance of sharing data."

This led the group to choose the open source PostgreSQL database over Oracle and Ingres and the PHP scripting language. "PostgreSQL is doing quite well as we now have more than 1000 images and 1.2TB of data," he said.

The Howard Florey Institute has now put in an application for an Australian Research Council grant valued at up to $200,000 to expand its computational resources.