This page contains a Flash digital edition of a book.
towards exascale

From mobile phone to supercomputer?

Tom Wilkie looks at the emerging strategies for

Exascale computing I

n early February, just aſter this issue of Scientific Computing World will have gone to press, the US Department of Energy will send to Congress its plan for building

an Exascale computer before the end of this decade. Japan too is expected to outline its Exascale plans within the next few months. In Europe, several Exascale-related projects are already underway, including an attempt to show that the most powerful computers in the world can be built using the processor chips from mobile phones. Momentum is clearly building, neatly

summed up by Jacques Philouze, from the soſtware company Allinea: ‘Exascale really is the future – it creates the buzz.’ Power consumption dominates every discussion of Exascale. Today’s Petascale

machines consume somewhere around 5MW to 10MW of power annually, costing between $5M and $10M at current US prices. Exascale machines will run a thousand times faster, so no-one can afford simply to scale up existing technology for no-one could face annual electricity bills of more than $5 billion. For Alex Ramirez, head of Europe’s Mont-

Blanc project, this issue of power consumption means revolution, not an evolution in technology. Exascale, he said, ‘will not be achieved by adding more of what we already have. We can see fundamental limits in power density and that means that many of the tricks of increasing the number of nodes are not going to work. ‘Instead of trying to take current

technologies and increasing their energy efficiencies, we are looking at the most energy-efficient technology today and trying to aggregate them.’ In a move that he freely accepts some observers might regard as ‘speculative’, the system architecture will be based on processors from ARM that

are currently used in mobile phones and embedded applications ‘because these guys have been facing power-density limitations from the beginning – they work with battery- operated devices where energy consumption was always an issue.’ But Ramirez does not believe that it will

be enough simply to improve the energy- efficiency of the processors alone, he is concerned about maintaining the ‘system balance’. He points out that processors account for only about 30 to 50 per cent of the total energy consumption, so the Mont- Blanc project is also looking at more efficient interconnects, memory and storage. British company Gnodal is supplying the

interconnects and here again, the technology is highly innovative. According to Tony Ford, COO of Gnodal, the company’s role is to integrate the low power processors more closely by providing very high-capacity interconnects between what will be very high- density compute resources. Although current Ethernet connections operate at 1 Gigabit,


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  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48
Produced with Yudu -