Exascale: The Beginning of the Great HPC Disruption
by Mike Bernhardt
Sketch of a helicopter-like flying machine by Leonardo da Vinci, c. 1483. Imagine how disruptive this would have been had it worked!
“It’s not the technology of Exascale that is disruptive, but the things Exascale makes possible that will be disruptive,” says Thomas Thurston, CEO of Growth Science International, “Exascale isn’t a 'disruptor,' it is a foundation upon which a million disruptions can be based. It enables others to be disruptive in ways and at a scale never before imagined.”
With the eventual arrival of exascale systems, we face a level of disruption that is unlike anything this community has ever experienced. But what are we really talking about when we throw these business pop culture terms around like Frisbees™?
As we look at the potential disruptive impact of exascale computing, we have to keep in mind that we’re looking at much more than bigger and faster computers or innovative new technology.
While much of the early exascale discussions will inevitably focus on system size, numbers of processors, power consumption, new programming models, and so forth, the real disruption to be considered is the impact exascale-levels of computation will have on bringing more capabilities, new discovery, and more efficiencies to a much larger number of people than ever before.
No Hoof, No Horse: The Long, Winding Trail to Exascale Storage
by John Kirkley
Study of a horse by Leonardo da Vinci. "For want of a nail..."
There’s a saying in equestrian circles, “No hoof, no horse.” In the world of HPC, this translates as “No storage, no system.”
In the not too distant past, when configuring a new supercomputer, storage was often an afterthought — an uninteresting but useful appendage that performed quietly in the background while rows of fancifully painted compute cabinets hogged the limelight. In one apocryphal industry story, a university alumni fund presented its alma mater with a leading-edge HPC system so it could join the hallowed ranks of the TOP500. The machine screamed through the LINPACK tests. Only problem was that after the handshaking and backslapping was over, the university’s researchers found they couldn’t use the system — no one had made any provision for adequate, high-speed storage.
More Than Just the Iron
by John West
Giant crossbow designed by Leonardo da Vinci, 1478, Codex Altanticus. The structure is huge, measuring 27 yards across.
Machines at 10, 20, and 50 petaFLOPS are large enough to break just about everything in the computing ecosystem. Hardware aggregated at this scale starts to show increased failure rates that we have only worried about theoretically in the past. Interconnects and system software have never been tested at this scale, and both hard and soft errors take flight from previously uninteresting corners of the system. And users begin to dig deeply into both their applications and the science supporting those applications as they attempt to break down the next important barriers in science. All of these stressors become more pronounced the larger the machine, and most people anticipate that the transition into exascale will be substantially disruptive to both users and to the centers fielding them.
Successful deployments need to manage these stressors for both their users and their programs, and the job is not easy.
Intel's European Investment
by Mike Bernhardt
In case you're counting, Intel now has three exascale labs in Europe. These research centers are all members of the Intel Labs Europe network:
- Exascale Computing Research Center, Paris (France). Established December 2009
- ExaCluster Laboratory, Juelich (Germany). Established May 2010
- ExaScience Lab, Leuven (Belgium). Established June 2010
Linked: Exascale in the DOE
by John West
The United States Department of Energy is the leading organization pursuing the development of practical exascale computers in the United States — no other US organization even comes close in terms of planning for the massive technology shifts we’ll need to get to exascale within a 20MW power budget by the end of this decade. And they are putting their money where their plans are, funding much of the needed R&D.
DOE Awards $3M for Component Research
by John West
A team from Pennsylvania State Uni-versity, Oak Ridge National Lab, Hewlett-Packard (HP), and the University of Michigan recently received a $3M (USD) award from the US Department of Energy for work on a three-year project to examine advanced architectures and critical technologies for exascale computing. The award is part of an effort managed jointly by the DOE and DARPA to explore approaches for dealing with key barriers on the path to exascale computing.
EU Searches for New 'Super Model' in TEXT Project
by Bob Feldman
Heard over a pint…
Richard Dracott, formerly the GM of Intel's High Performance Computing organization, will be taking a temporary position as the Acting Director of Intel's European exascale labs. We'll catch up with Richard in the near future to get an update from him.
IBM is turning up the heat on hiring for exascale. In the most recent post we saw, they are looking to fill a Ph.D. position in exascale computing at the Heterogeneous Computing Laboratory in Ireland. The lab is located at University College Dublin (UCD), and the position is co-funded by the Irish Research Council for Science, Engineering and Technology (IRCSET). As an interesting note, the ad for this position states that the exascale activity at the lab will adopt a Stream Computing paradigm and will use streaming hardware and techniques.
Join the Discussion
The Exascale Report encourages members of the HPC community to contribute articles dealing with key issues and technologies associated with the drive to develop exascale systems. Please submit a brief (150-250 word abstract) to idea@theexascalereport.com describing your proposed topic. We welcome original submissions from HPC practitioners. Articles will be peer reviewed.
Please note that success stories, white papers, press releases and other marketing communications collateral or public relations materials are not what we are looking for.