Where do aged supercomputers go to die? New Mexico
Saturday, January 5th, 2013Moore’s Law puts supercomputers out to pasture since energy – not usually a cost of electricity, though a accessibility of extract – is a biggest imprisonment during a large supercomputing centers. And infrequently a miss of bill helps close a gate, and HPC cloud computing butchers a cow.
That’s a box with a large 28-rack supercomputer during a University of New Mexico nicknamed “Encanto,” that was ostensible to be a substructure for high-tech investment and investigate in a Southwest state with a sign “Land of Enchantment.”
The thought with a Encanto plan was to lease out ability on a box, as El Reg reported when a Xeon-based cluster, built by Silicon Graphics regulating 3,500 of Intel’s four-core Xeon processors and indeed using in an Intel information center, was initial dismissed up.
The Encanto plan was allocated $14m by a New Mexico state supervision in 2007, with $11m of that going for a complement itself and $3m going for gateways – presumably a brew of networking rigging and telecom services – to couple into a complement from around a state. At a time Encanto was initial incited on, it cost $2m a year to keep it running. So over 5 years, it cost as many to run a supercomputer as to buy it.
The Encanto Altix ICE cluster now has 14,336 cores, a rise fanciful opening of 172 teraflops, and a postulated opening or 133.2 teraflops on a Linpack Fortran benchmark. That gives it a ranking of 185 on a many new Top500 supercomputers list from Nov 2012. The largest machines have between 50 and 100 times some-more oomph than Encanto, and also large budgets from a incomparable inhabitant or transnational supercomputer organizations that can means to compensate and run these grievous machines.
According to a report in a Albuquerque Journal, New Mexico state information record secretary Darryl Ackley is going to clout adult Encanto into pieces and put it out to pasture in several investigate universities in a state. No one wants to buy Encanto as-is since of a high cost of using a machine.
The irony is that 6 years ago, when Encanto was being conceived, a Altix ICE 8200 machines were pretty power-efficient. Since then, however, Intel, AMD, and IBM have cranked adult a core count and lowered a thermal envelopes of their floppiest processors, assisting machines bake significantly reduction juice.
Encanto browns 861 kilowatts, that works out to 154.7 megaflops per watt postulated performance. The 17.6 petaflops hybrid CPU-GPU “Titan” supercomputer during Oak Ridge National Laboratory browns 8.21 megawatts, delivering 2,142 megaflops per watt. That’s opening per watt alleviation by a cause of about 13.8.
The irony is that we are going to need a many incomparable jump in potency to get to exascale computing in a 25 megawatt energy budget, as a vital HPC players are operative together to accomplish by 2018 to 2020. That’s a cause of 50 times a opening boost in usually 3 times a energy bill to get around 40 gigaflops per watt in something on a sequence of 200 to 300 racks.
New Mexico’s Encanto supercomputer
The Encanto appurtenance was ostensible to have rented out adequate ability to New Mexico’s universities, industrial companies, and consultants to be self-sufficient in 5 years, and as El Reg forked out 5 years ago, this all had a pie-in-the-sky feel to it, and Encanto would be a rather run-of-the-mill supercomputer by 2013. And so it was, and so it is.
Ackley told a Albuquerque Journal that a few tire-kickers came looking during Encanto, saying if they competence wish to buy it, though it was too costly to operate. The University of Mexico is slated to take 10 racks of a box, with New Mexico State University holding 4 and a New Mexico Institute of Mining and Technology holding two.
If a machines are giveaway or have a medium price, afterwards this creates sense. Otherwise, those schools should simply buy a complicated Xeon or Opteron cluster – maybe with Tesla GPU or Xeon Phi coprocessors – and stop vital in a past.
The doctrine here is that if we wish to build a supercomputer estimate and focus growth center, your bill has to be sufficient to not usually run a supercomputer though also to ascent it and make it appealing for complicated applications.
As usual, it takes income to make money. And for a lot of companies, shopping some ability on Amazon creates a lot some-more clarity and is substantially easier – and cheaper – than whatever New Mexico could come adult with for Encanto. Moore’s Law competence have killed Encanto, though AWS is nailing a coffin shut. ®
Article source: http://www.theregister.co.uk/2013/01/04/encanto_super_chop_shop/
Data core managers and CIOs face a flourishing plea when it comes to datacenter efficiency. With surging electricity costs and arise in information expenditure rates in organizations, potency is apropos a vital priority in a information center. CIOs have now begun to comprehend that information storage can no longer be put on a behind burner and therefore in 2013, a concentration will be on improving a potency of their datacentres so as to get a best value from their investments. Analysts trust that use of fit inner IT approaches like virtualisation, software-defined networks (SDNs) and a use of converged infrastructure solutions can assistance solve this vital challenge.
