Skip Navigation | ANU Home | Search ANU | Directories

Search RSES

---

TerraWulf II - Overview

---

 

Project description:

TerraWulf is a networked ‘Beowulf’ cluster of computers set up at RSES to provide convenient high end computing power for a range of demanding geoscience problems, particularly in the fields of thermochronology, seismology, geodesy crustal and mantle dynamics, and landform evolution.

The main cluster, TerrraWulf II, consists of 96 nodes (384 processor cores) connected through Gigabit and Infiniband switches which significantly extend the range of potential applications of the cluster to both `tight’ and `loosely coupled’ codes.

The new Cluster was ordered in mid 2007 and delivered in August. Upgrades to the power and cooling infrastructure continued too the end of the year.

The TerraWulf II cluster will be integrated into the AuScope grid and used for simulation and analysis of terrestrial processes and variety of geoscience data processing as well as continuing development of inversion and data inference software.

Background:

In 2003 we developed a Beowulf cluster as a platform for solving complex data inference problems in the Earth Sciences, and in particular the fields of thermochronology, seismology, crustal and mantle dynamics, and landform evolution. A Beowulf cluster is a network-linked set of commonly available `off-the-shelf' computers configured to give high performance/cost ratio. Projects using the TerraWulf facility combined state-of-the-art computational techniques developed at ANU with high quality data sets collected over the previous decade to address fundamental questions in the Geosciences.

A major upgrade of the TerraWulf compute cluster was recently made possible by funding support from the National Collaborative Research Infrastructure Strategy (NCRIS) under the AuScope umbrella.  Contributions from the AuScope Geospatial, Imaging and Structure, and Access and Interoperability components combined with support from RSES has enabled the construction of a new and more powerful cluster.