Racing Down the Long and Winding Road to Exascale

Print Friendly, PDF & Email
As we glance at 2011 in the rear view mirror, it’s hard to believe that The Exascale Report has been publishing for eighteen months. When we started our subscription-based publication to focus on a topic receiving very little coverage at the time, we really had no idea what kind of acceptance we’d find. Today we are going strong with a growing international base of readers. We’re proud to be playing a small role in helping bring together the future exascale community.

With such a long way to go between now and the exascale target timeframe of 2020, many in the industry have understandably described the industry’s efforts to move forward as a journey. The journey has now officially turned into a race, and those who have formally entered the race include China, Japan, Europe, the U.S., Russia and India.

For many years, technology leadership, particularly around HPC, seemed to be entitled to the U.S. But not anymore. False confidence, which some have described as U.S. arrogance, political infighting among U.S. funding sources, and perhaps even a lack of belief or understanding of the importance of a national science and education foundation have all been factors in how this race is shaping up.

A “Slow and Steady” approach as depicted in our political cartoon from 1864 (predicting the victory of incumbent U.S. President Abraham Lincoln over his rival George B. McClellan) is not going to win the exascale race. This is a race in which we need to pull out all the stops.

If Only We Had a Roadmap.

With so much activity taking place under the umbrella of exascale, it seems like very little has actually changed. And despite some company’s claims that they know how to get us to exascale, there are still many unanswered questions on the technology front.

Some countries seem to have a better vision. While concerted efforts in several countries have produced rather impressive early results, the U.S. continues to stumble along the road, with key influencers often arguing at every turn as to which path to take. Pressure to drive commercial product revenue by the industrial sector is creating a tradeoff game that pits short term revenue generation needs against longer-term, higher risk research necessary to produce breakthrough technologies.

This point comes out in several of our featured interviews and is presented in a nicely balanced perspective from well known and widely respected HPC luminary, Professor Thomas Sterling, in his Exascale Report exclusive contributed article titled, “In Search of an Exascale Roadmap.”

As you will hear from the all industry luminaries who have contributed to this issue of our report, there is a sense of urgency regarding funding and collaboration if the U.S. is going to compete in this race. The sponsor, and often the pit crew for HPC innovation in the past, DARPA, is out of the race and the responsibility for driving the U.S. HPC roadmap is now resting with the U.S. Department of Energy.

We can’t build a roadmap solely based on the technology we understand today, and we can’t win a race without breakthrough technology in several areas. Other countries are outspending us significantly and the somewhat surprising change in global technology leadership is the result. If the U.S. doesn’t wake up and make the necessary investments in 2012, they could be out of the race for good.

Pretty much everyone we’ve spoken with feels the leadership of this effort at DOE is in good hands with Dr. William (Bill) Harrod, a former DARPA program manager and a real HPC industry veteran. I think it’s fair to say concerns are not with the leadership, but with the lack of support and adequate budget being given to Harrod.

[ Dr. William Harrod is currently the Division Director for the Department of Energy/Office of Science Advanced Scientific Computing Research program’s Computational Science Research and Partnerships Division.]

Of the many interviews we conducted in November and December, we’ve narrowed down this month’s editorial to only four. The four contributors to our final issue of 2011 are experienced and respected spokespeople for this community as well as stakeholders in the advancement of HPC. Thomas Sterling, Pete Beckman, Wilf Pinfold and Sumit Gupta share their opinions, their enthusiasm and their concerns as we enter a pivotal year for exascale R&D.

2012 is indeed a pivotal year. General industry consensus is that progress in 2011 has not been so good, despite many hyped up claims. (See our 2011 Exascale Progress Meter)

To sum this up, something is not right with this race, even for those who are cautiously optimistic. Speaking now as a U.S. patriot, I have to ask if we are racing down the road with blinders on. Maybe we’re inhaling the exhaust fumes of those leading the race and it’s causing us to not think clearly. As we attempt to race down the track at a breakneck speed, without a clear roadmap, the likelihood increases that we’ll spin out of control and slam into a wall.

In fact, if the U.S. doesn’t step up its game – the writing may already be on the wall.

Mike Berndardt

Comments

  1. exa-preview-1 says
  2. exa-preview-1 says
  3. exa-preview-1 says