This page contains a Flash digital edition of a book.
HPC 2013-14 | Energy efficiency

could potentially get more performance per financial investment as they can simply buy a lot more of them. Once all the background work has been

completed, a power model and a performance model will be built and then integrated. Te focus will then shiſt to how different implementations of the same algorithms behave in terms of power usage. Two very common applications – a simplistic code that performs a very important mathematical operation in relation to seismology modelling, and a much larger Lattice Boltzmann fluid dynamics code owned by EPCC – are among a set of applications that have been chosen as test cases. By developing a tool that can model the performance and power usage of typical scientific soſtware on any type of hardware available, the project members hope to open up the possibilities in terms of hardware procurement and enable users to modify their codes in order to reduce power consumption.

Pay by the power? In conclusion, it’s important to note that hardware vendors are working with soſtware vendors to ensure that efficiency doesn’t come at the expense of performance, and vice versa.

The Convey hybrid-core system delivers increased application performance with lower costs

Engineers from SGI and Altair, for example, began collaborating in 2012 to produce an energy-aware computing solution called ‘performance power resilience scheduling’ (PPRS). Tis solution marries Altair’s PBS Professional, an HPC workload management and job scheduling tool, with the SGI Management Center. Te resulting technology enables administrators to create power profiles which are associated to individual jobs that will be executed within the cluster. Scott Suchyta, director of partner solutions

and integration at Altair, explained: ‘If a certain type of job can run with 20 per cent less power

but still maintain performance, then each job that falls into that category will be assigned that same power profile. Also, the workload manager will forecast to identify available nodes and prioritise work accordingly.’ He added that, by throttling down the power consumption of certain jobs, significant power savings can be made with only a slight impact on the time it takes the job to run. Te PPRS solution has a target release date of the end of 2013. Moving beyond technological advances

and optimisations, a mind shiſt is what’s really needed in the industry, said Suchyta. ‘My hope is that we’ll see a day when users are charged not for the CPU hours they use, but for the amount of kilowatt hour their jobs consume,’ he said. ‘Tis mentality will allow users to think

more about what they need to do to get their job to run as quickly as possible, using the least amount of power. If we could get over that hurdle of users being focused purely on time to results, the whole industry from data centre operators to soſtware developers would be focused on the same goal and we could make real progress in the area of energy efficiency.’ l

Direct Contact Liquid Cooling

Industry leading data center solutions for the HPC, Cloud and Enterprise markets.

The RACK DCLC™ product line provides scalability, modularity and fl exibility to meet your density, effi ciency and performance requirements. Demand LIQUID

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36