ACADEMIA

Bright Computing announced that Massey University’s department of Theoretical Chemistry and Physics is now using Bright Cluster Manager to manage their powerful HPC system. The department switched from Rocks and other open source toolkits to Bright as they expand their cluster for immediate needs, and as they plan their future hardware evolution. As a result of this change, Massey University realized immediate gains in operational efficiency and reduced system management workloads. Bright also vastly reduced the time and effort required to install and test new hardware as they continually evolve the system, maximizing available compute time.

Massey University’s department of Theoretical Chemistry and Physics comprises 20 scientific staff conducting more than 30 diverse research projects on the far edge of science. Their workhorse for these research projects is a 624 CPU core, 2TB RAM, and 100TB disk cluster located on campus.

The range of research at Massey is impressive. For example, one team is investigating whether the speed of light and other fundamental constants are really constant over space and time; and if not, determining how they change. This project involves analyzing the spectra from distant quasars to determine whether the speed of light has changed over the billions of years since it was emitted. Very highly accurate calculations of the dependence of atomic and molecular spectra on the speed of light are necessary for this research. In addition, the researchers are working to discover molecular systems that would have an enhanced sensitivity to the change in fundamental constants, potentially allowing experiments to take place on Earth.

Scientists at Massey are also important contributors to the study of super-heavy elements: atoms that don't exist in nature, but are artificially created in particle accelerators. The actual experiments are only possible in a few locations on earth – and cost millions of dollars to perform. Massey scientists use HPC to predict atomic and chemical properties of these elements. They also help the researchers determine the feasibility of experiments, planning the research, and interpreting the results. These calculations are extremely compute-intensive, requiring very efficient HPC systems.

A third example of Massey’s research is a project on the frontier of nano-scale physics. “Nano-devices are small enough that chemistry and quantum theory play important roles, but large enough that accurate quantum theoretical simulation is a staggering task,” said Dr. James Avery. “This necessitates inventing new computational methods and performing massive calculations to understand these materials, and to predict the strange ways in which they behave, ”

The diverse nature of research conducted at Massey creates a wide array of unique system requirements for their cluster. In the past, setting up the cluster for these jobs consumed seemingly endless hours of system administrator time, driving down overall system throughput whilst burdening talented researchers with tedious tasks. Now, Bright’s image-based provisioning enables the scientists to reconfigure their cluster to for the specific demands of each project in minutes, vastly increasing productivity and freeing manpower for other priorities.

The team at Massey realized other benefits from using Bright.

“As we looked to evolve our system to meet our expanding needs, we were driven to find a better way to run our cluster. We are scientists; we want to spend our time on science. Not on provisioning, monitoring and management,” said Dr. Michael Wormit. “We were using Rocks and other open source toolkits, wasting far too much time on customization and keeping the various tool versions synchronized. We were reluctant to make changes to the system — it created too much overhead.”

“What appealed to us about Bright Cluster Manager is that it is fully integrated and easy to use,” said Dr. James Avery. “It’s leaps and bounds better than anything we have worked with in the past.”

Using Bright has enabled the department’s three part-time system administrators to shift their workloads to priorities that were previously put on hold due to day-to-day system administration demands. Now they are able to work on initiatives to improve their cluster’s performance, such as setting up InfiniBand to reduce latency.

“Installing a cluster with Bright is much easier and faster,” added Dr. Michael Wormit. “It’s especially useful that we can install new compute nodes to our cluster in minutes, or quickly re-purpose hardware. Bright minimizes the effort of tasks that previously took a lot of work, in a straightforward, intuitive manner. It’s also helpful to us that Bright has a complete development environment with everything we need.”