Evolution or Revolution?
by John West
Thomas Sterling and Pete Beckman argue the "revolutionary" and "evolutionary" paths for achieving exascale computation by the end of this decade.
Over the past 60 years HPC has changed in dramatic spurts punctuated by fairly long periods of stability. During that time we've been through five computing epochs in HPC, including sequential execution, sequential issue, vector, and SIMD models of computation processing. And along the way there have been excursions into dataflow, systolic, and global shared memory models.
Each of the phase changes from one model to another was precipitated by a prior change in hardware design and system architecture. Often our programming models were a consequence of the hardware designer's art, not the result of a collaborative effort between software and hardware designers to develop the most sensible system.
Today we stand firmly in the epoch of communicating sequential processes, enabled primarily by MPI. As we look forward to achieving the goal of a sustained exaFLOPS system by the end of this decade, we asked two leading experts in the exascale community this question: are we on the cusp of a new phase change, a new revolution, in HPC, or can we extend and adapt today's programming model to get us there?
Can the Exascale Effort Survive the Need for Global Cooperation?
by Mike Bernhardt
The Battle of San Romano. Paolo Uccello, c. 1438-1440.
Technology leadership has long driven fierce competition among nations.
The Top 500 list of the most powerful computers in the world has evolved into a measurement not just of system benchmark performance, but one of bragging rights to claim technical prowess for system manufacturers and the countries where these systems are installed. Having a system placed on the Top 500 has become an annual quest.
Now, a new quest is underway — the quest to achieve exascale-level computation. Industry experts say we'll never get there without many years of global cooperation. Visionaries say this is not a problem, the global HPC community will come together to achieve this dreamed of goal. But naysayers contend that global cooperation sounds good — and it might even work for the short term — but it's more likely that the global cooperation will turn into fierce global competition as exascale moves closer to reality.
DARPA Jump Starts a Revolution
by John Kirkley
Bohumil Kubišta, Pobrezni guns in combat with the navy (1913).
Anyone familiar with DARPA knows the agency is not averse to taking risks. Tackling really tough technological problems by funding innovative research is fundamental to its mission statement.
But with the Ubiquitous High Performance Computing (UHPC) program, DARPA is really pushing the envelope. It’s calling for nothing less than a revolution that could drastically and permanently alter the nature of computing.
Response from the HPC community has been predictable. Two opposing camps have sprung up — those who side with DARPA’s call for a totally new computing paradigm versus those who believe the road to exascale should be evolutionary — leveraging what is already in place, moving along the roadmap. But Bill Harrod, DARPA Program Manager for UHPC, makes no bones about where he and the agency stand. As far as the BAA is concerned, incrementalists need not apply.
DARPA Pushes High End Computing Frontier
by John West
On June 21 DARPA announced the latest in its recent spate of high-end computing efforts, the Omnipresent High Performance Computing (OHPC) program. This effort will award investigators up to a maximum of $3M over three years to build hardware that will be needed by the systems activities undertaken as part of DARPA’s Ubiquitous High Performance Computing (UHPC) program.
EU Considers New Exascale Initiative
by John West
In mid-June members of the Information and Communication Technologies component of the European Union's Seventh Framework Programme (FP7) discussed plans in the works to add a focus area for exascale technologies to their research program.
In a presentation given in Brussels on June 11 Leonardo Flores Añover outlined the proposed exascale objectives in FP7. If adopted this would mark the first objective in FP7 aimed at the exascale, complementing other EU-funded efforts such as PRACE and support for the International Exascale Software Project.
New Ethernet Standards Hold Promise for Big Science
by John West
The conversation about exascale often focuses on the computation and on getting to a computer system capable of an exaFLOPS rate of computation. But the data created by and used within those computations has to get in and out of whatever computers we end up building, and today's networks are not up to that challenge. In late June the IEEE Standards Association announced the ratification of the new IEEE 802.3ba standard which includes both 40Gb/s and 100Gb/s Ethernet, and the scientific community is already moving to adopt the new technology.
Exascale Development Flows to Europe
by Bob Feldman
Intel and ParTec agreed to create the ExaCluster Laboratory. From left, Achim Bachem, Kirk Skaugen, Hugo R. Falter. (courtesy Intel)
DOE to Collaborate with Cray on Next Gen Interconnect
by Mike Bernhardt
Los Alamos and Sandia National Laboratories have formed a new HPC center called The Alliance for Computing at the Extreme Scale (ACES). This collaboration calls for the two labs to jointly architect, develop, procure and operate capability systems for DOE's Advanced Simulation and Computing Program (ACS), according to a paper presented at the 2010 Cray User Group meeting this past month.
On the Web
The International Exascale Software Project is the exemplar of the kinds of global cooperation our community will need to sustain if we are to build a usable system capable of delivering an exaFLOPS of computation by the end of this decade.
Happily, the IESP is also an exemplar of open communication. Briefings, reports, and summaries from all four meetings, including most of the presentations given over the past year, are available on their web site.