Linux Clusters at NIST

NIST is using Linux clusters for research, benchmarking them against supercomputers.
Conclusion

Linux has been beneficial in our research. The first device driver for the PCI MultiKron card was done on Linux and was the easiest to write. We use Linux to monitor the cluster, and the tools we develop are either written for Linux first or ported quickly from other UNIX environments. Experimenting with computing clusters would be more difficult with commercial operating systems because source code is generally not available. By having the ability to probe the operating system source code, we are able to accurately measure performance of the OS in addition to the performance of our applications.

Our experiments show that clusters compete very well with traditional parallel machines when running distributed memory applications, generally characterized by large messages. For shared memory applications, which tend to communicate with many small messages, the overhead of the network has a detrimental effect on the application performance. For both types of applications, tuning the network parameters can be of tremendous benefit in decreasing execution time.

The 333 MHz, 16-node Pentium-II cluster has been transferred into a production environment. This cluster will be made available to the entire NIST community and will be managed by the group that supports the traditional supercomputers. We believe Linux-based clusters will provide an effective environment for running many high-performance applications.

Resources

Acknowledgements

Wayne Salamon is a Computer Scientist within the Information Technology Laboratory at the National Institute of Standards and Technology in Gaithersburg, MD. He has worked on system software for PCs, UNIX workstations and IBM mainframes for the past 12 years. His current research interests are parallel computing and performance measurement. Wayne can be reached at wsalamon@nist.gov.

Alan Mink is project engineer of the Distributed Systems Technology project within the NIST Information Technology Laboratory. He holds a B.S. in Electrical Engineering from Rutgers University and an M.S. and Ph.D. in Electrical Engineering from the University of Maryland. His research interests include computer architecture and performance measurement. Alan can be reached at amink@nist.gov.

______________________

Comments

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

To learn more about ATM Software

Thomas_Stevens's picture

Hey Guys,
Great article here. I was wondering if you had any experience with Diebold's Agilis ATM Software or it that was a consideration in your metrics here.

I found this link:
ATM Software

But wasn't sure how much a consideration that was in your research. Any insight would be appreciated. Thanks,
Thomas

Geek Guide
The DevOps Toolbox

Tools and Technologies for Scale and Reliability
by Linux Journal Editor Bill Childers

Get your free copy today

Sponsored by IBM

Upcoming Webinar
8 Signs You're Beyond Cron

Scheduling Crontabs With an Enterprise Scheduler
11am CDT, April 29th
Moderated by Linux Journal Contributor Mike Diehl

Sign up now

Sponsored by Skybot