Research highlights the future of energy-aware high-performance computing

Leipzig, June 18, 2014. As reported in Inside HPC.

With energy costs a growing concern for High Performance Computing, Allinea Software will demonstrate its vision of a greener future with a preview of new tool extensions for application energy usage optimization at this year's International Supercomputing Conference (ISC'14) in Leipzig.

With larger numbers of data centers consuming over 1MW of power or having electricity bills topping $1M, energy is focussing the minds of the system sponsors and managers, says David Lecomber, CEO of Allinea Software.

Allinea Software worked with application performance experts at the University of Warwick to investigate novel energy and power measuring techniques for scientific application workloads.

Energy usage data is increasingly available at the system level, and our research also explored proxies for energy such as hardware counters to see where they could give deeper insight, said Professor Stephen Jarvis of University of Warwick’s HPC Performance Analysis Group.

The research supported the view that in many cases applications can reduce energy costs without adversely impacting actual run time.

Improving the green credentials of hardware and data centers is vital, and progress is good, but applications must also play their part. Energy optimization is a natural fit for our performance tools, adds Lecomber.

With the variety of workloads that HPC centers have, a ‘one size fits all’ strategy is a costly error – and so Allinea Performance Reports will provide information for application users and system managers to enable them to tune system and application parameters such as CPU frequency for optimal energy use for each application.

Our developer-centric tool, Allinea MAP, will allow scientific code developers to focus energy optimization down into the source code – making changes to the application to drive faster performance and lower energy consumption at the same time.

This research has been supported by the Technology Strategy Board's Emerging Technologies Energy Efficient Computing Programme.