Brecca High Performance Computer at Monash

The VPAC 'Brecca' High Performance Computing (HPC) facility was relocated to Monash in mid 2007, following completion of its primary four-year run life at VPAC.  Subsequently the small Monash School of Mathematics 'Sputnik' cluster was decommissioned, adding a further 16 CPUs to Brecca. The combined facility comprised 196 CPUs of which Monash had access to 62.5% (around 122 CPUs). The total theoretical capacity of its 196 Intel Xeon P4 CPUs and around 340 GB of RAM was approximately 500 Gflop/s. The facility significantly increased the HPC capacity available at Monash at the time, and the remainder of its capacity was available to other VPAC users and for other general purpose VPAC resources.

Brecca was decommissioned on 9 July 2009, following completion of a two-year run life at Monash. Before that, the Centre worked with Brecca users to migrate their work onto significantly higher performance and higher efficiency HPC facilities which were made available at Monash to replace Brecca. As a part of the migration process, Brecca was slowly reduced in capacity as components progressively failed, as it was well beyond its manufacturer’s support life.

Using Brecca

In its reincarnation at Monash, Brecca was configured as a 'grid' resource. This means that instead of having a login account on Brecca, users submitted jobs to Brecca using Globus.

Information about accessing Brecca from Monash is available here.

Load and utilisation of Brecca at Monash is available at:
    http://www.vpac.org/supercomputers/usage/4/
    http://status.arcs.org.au/hosts/ng2.vpac.monash.edu.au/

Further information about Brecca is available from VPAC.

Showcase

During its first full year of operation at Monash, Brecca had already contributed 700 thousand CPU hours of valuable HPC compute time to Monash and other research programmes.

	Bioinformatics in Protein Crystallography, Steve Androulakis and Dr Ashley Buckle, Department of Biochemistry and Molecular Biology This project uses bioinformatics to assist in protein crystal structure determination, by comparing new structures against the entire Protein Data Bank library, which comprises over 70,000 structures. To do this, the research team uses the Nimrod parametric sweep job dispatcher to run CPU-intensive, embarrassingly-parallel molecular replacement operations using the Phaser software tool running on grid HPC resources including Brecca.
	Latin Squares, Dr Ian Wanless, School of Mathematical Sciences. This project aims to develop the theory on important substructures of 'latin squares'. Latin squares play an important role in pure mathematics and have many applications in the areas of combinatorial design, such as event programme scheduling, statistical experiments and error correcting codes. The research team uses distributed computing including the Brecca HPC as an investigative tool to find patterns which can then be turned into mathematical theorems by humans.
	Molecular Dynamics Simulation, Cyril Reboul, Department of Biochemistry and Molecular Biology This project uses molecular dynamics (MD) simulations to investigate the mechanistic molecular details of GAD, an essential enzyme involved in the production of a major neurotransmitter inhibitor. The research team uses the Brecca HPC to perform the MD simulations which are computationally intensive and can only be run on large scale parallel computer clusters.