![]()
|
The Academic Computing Services High-Performance Computing InitiativeBy Dr. Philip Baczewski, Director of Academic Computing and User ServicesSince 2001, UNT's Academic Computing Services has supported multiple clusters of 32 and 64-bit processor systems running Linux for compute-intensive scientific research. Cluster computing provides dedicated systems for concurrent processing of jobs in a batch environment. Most jobs run on a single compute node and can utilize up to 100% of that node's processor and memory resources. Additionally, we support a cluster in which multiple nodes can be used to run a single computer program that has been optimized for this kind of parallel processing. Recently, we have added two high-performance multi-processor systems that can run jobs previously possible only at national super-computing centers. These high-performance systems have been used primarily for scientific research, however, they are available for any computationally-intensive research done by UNT Faculty and graduate students. Our current configuration of systems includes 50 32-bit CPUs, 104 64-bit CPU cores, 336 Gigabytes of RAM, and about 20 Terabytes of disk storage. This information is not a useful technical description, but it does provide an interesting overview of the facility. The following table shows the current configuration of our HPC systems.
*Standalone multi-processor systems
**Parallel processing capability via a Myrinet high-speed interconnect between nodes ***Storage listed is for cluster controller systems The following graphs show usage trends for the ACS HPC systems. The graph immediately below shows the increase in the number of compute hours on an annual basis. In 2005 alone, that's about 28.5 years of compute time used on all HPC systems combined.
This next graph shows the research areas which have utilized the ACS HPC systems in the 2005 calendar year, partitioned by hours of compute time. Our greatest amount of usage has been in the area of computational chemistry, however, math and other areas have employed HPC systems for their research problems, including a grant-supported project in political science.
|