Powerful computing clusters which drive up firms' costs due to energy-sapping multi-core applications could be about to get slicker thanks to a grant from the European Commission to develop load-balancing technologies for industry.
Firms reliant on high-performance computers or embedded systems could benefit from a £2.9 million (€4 million) European Commission grant to research and develop new ways of balancing loads on complex, multicore applications.
What's the problem?
Computers with multiple processing cores have become the norm. They are found in smartphones, tablets, game consoles on the smaller side of the spectrum, through to PCs and servers.
As technology develops the numbers of cores for every device will too, which spells problems with load balancing. Poor load balance is the difference between a phone battery lasting for hours rather than days, for example. On a larger scale, cloud computing clusters with poor load balancing can cost firms - and the environment - dearly.
A group of technology developers, industrial systems suppliers and research organisations have joined forces to optimise computing resources for computing clusters often seen in automotive, media, scientific and pharmaceutical industries.
Bosch, York University (who will lead the technology research) and leading supercomputing facility HLRS are a few of the firms to sign up to the ‘DreamCloud’ project.
It is hoped that ‘DreamCloud’ will develop systems that can provide the most powerful computing with the most efficient use of energy.
Dr Leandro Soares Indrusiak, technical lead at University of York, said: “What’s needed for the next generation of applications that seek to benefit from multicore platforms are new ways of dynamically adapting multicore processing to different application loads and energy constraints.
“We’re seeking to exploit today’s knowledge of biology, market dynamics and control and feedback theory to create new paradigms for the way processing resources are scheduled and dynamically adapted for complex application configurations.”
Different tools for different industries
The group will focus on developing three distinct types of cloud-like tools to manage application workloads on varying multicore platforms.
'Embedded Clouds' will be used for systems with time-critical behaviour, like the flight control in an aircraft, allowing for restricted load balancing and providing strict performance guarantees.
'Micro Clouds' will rely on new extensions to operating systems and virtual machines allowing for rapid migration of processing from one core to another.
Thirdly, 'High-Performance Clouds' will balance highly dynamic workloads, using the entire computing power of the underlying platform.
David Lounsbury, CTO of the Open Group consortium that are working on the project, said: “The DreamCloud project brings together experts in system software technologies that will develop new and creative approaches for balancing the load of multiple applications running across dozens or hundreds of cores while ensuring required levels of performance and energy usage.”
He added that the initiative could “hold tremendous promise for enabling new applications to be developed that more fully exploit the processing capabilities of next generation multicore platforms while providing the performance guarantees needed for critical systems in avionics, automotive, communications and other related domains.”