More Shameless Self-Aggrandizement

Quick note that a certain journal article has been covered by both International Science Grid This Week and HPCwire.

The HPCwire article is especially interesting as it calls us out for asking the tough questions. So be it -- HPC requires a balanced environment for productive science. You can't print a hundred terabyte dataset, nor can you build a hundred trillion pixel power wall. So if you need to build a machine that generates large datasets, you also need to plan for a post-processing environment capable of extracting the science. The exact mix of hardware, software, and skinny guys will depend on the site, but it's not something that good HPC sites ignore.

And yes, I realize that those questions are substantially less tough for me at a dot com, as compared to the co-authors who are at various dot edu and dot gov sites. Of course, that mix of employers also indicates that a lot of sites are thinking about this and "doing it right".

Local Boy Does Well

A quick note that the whitepaper version of an article I contributed to garnered mentions by Vizworld and InsideHPC last week.

Thanks to Wes for being leader and chief cat-herder -- getting all of us on a phone call is a herculean effort under normal circumstances, and it was even more challenging because I was in China working on Nebulae during the critical edits. It was an honor to be involved with a who's-who of HPC visualization.

Disruptive Technology

disrupt: to interrupt the normal course or unity of

- Merriam-Webster

With somewhat regular frequency, I get into discussions about whether GPUs are a "disruptive technology" in HPC (I hang out with nerds). Often this is in response to articles like this. As someone who now covers HPC for NVIDIA, but was working for LLNL through most of the last decade, I have some decidedly strong opinions on this topic.

The disruptive part is the upset in the top500 performance curve. If you plot the up-and-to-the-right performance of the top500 list, IBM BlueGene/L was above the curve as it existed at the time -- same outlay of cash/power/time, lots more performance. As the first massively parallel processing machine based on embedded processors, it was assumed that subsequent improvements in PPC (or other) embedded processors would provide a series of increasingly faster MPP systems. This has been mostly true, though the follow-on machines have not been large enough to capture the number 1 position.

Going back further in time, I don't think I ever heard the Earth Simulator referred to as disruptive technology. Though it captured the and held the "worlds fastest computer" title by a larger margin and for a longer time that BG/L would, most viewed it as the last gasp of vector technology. Distributed memory clusters like IBM SP had already won the mind-share at most sites, and though Earth Simulator was able to remain at #1 for over two years, it was pretty clear that it's dominance would come to an end. More vector machines were built, but HPC codes continued to move from vector to distributed memory.

Getting back to the question, with BlueGene/Q poised to re-take the top500 crown, are GPUs a temporary disturbance of embedded MPP march to dominance, or a permanent shift?

My answer is that disruption isn't a zero-sum game, and embedded MPP doesn't have to fail for GPUs to succeed. GPUs provide the "same slope, higher intercept" disruption of the top500 curve that embedded MPP did. Like embedded MPP, advances in the underlying technology will continue to provide GPU computing with performance improvements for years to come. And like embedded MPP, GPUs are leveraging a robust consumer market to achieve those advancements.

Where GPUs hold a significant advantage is barrier to entry. There are millions of CUDA-capable GPUs in systems today, and hundreds of universities teaching CUDA. Moreover, GPU clusters can be constructed entirely of common off-the-shelf hardware and software, putting the cost within reach of individual researchers and small teams. It's instructive to remember that Nebulae and Tianhe-1A were designed and built in months, while most of the embedded MPP designs have taken years to go from powerpoint to power-on.

There is room for both. In the upper range of the top500, embedded MPP will continue to leverage specialized networks and OS stacks to achieve performance through high node counts. At the same time, GPUs are already delivering systems with over 1TF of peak double-precision per node on everything from desktops to petaflop clusters. Both have been disruptive, and either could grow to dominate HPC in years to come. I've adjusted my career path accordingly for the next five years or so.

First Post: 2010 in Review

My name is Dale Southard. I'm a senior solution architect with NVIDIA, where I cover primarily high-performance computing and cloud computing. Results Over Coffee (ROC) is my blog, and I'll be posting on HPC and Cloud as well as some how-to information on programming GPUs, integrating them in Linux clusters, and whatever else I can cover over a cup of coffee. When I'm posting here, I'm speaking for me, not NVIDIA (even if I am posting about NVIDIA hardware or software).

I've often compared my work at NVIDIA to being a professional debugger -- I spend a lot of time working with very new hardware and software in new and complicated deployments. It's a lot of work, a lot of fun, and I feel lucky to be doing it for NVIDIA. In 2010, my personal high points were:

• Spent three weeks in China for debug/tuning the top500 runs on Nebulae
• Gave a talk at GTC on the above experience
• theRegister wrote a mostly-accurate article on above talk
• Spent a lot of time getting GPUs working in the Amazon EC2 cloud
• Spent a few days doing debug/tuning for the top500 runs on Keeneland
• Gave a presentation on GPU Cloud Computing at SuperComputing
• Had an interview with HPCintheCloud on GPUs and HPC (on cell phone, in traffic)
• Co-wrote an article on HPC visualization for IEEE with a lot of smart people
• And a bunch of other things that I can't talk about yet....

I have long viewed GPU Computing as a disruptive technology for high-performance computing and I am excited to have a ringside while it unfolds. ROC will be my place to share thoughts, tips, and hopefully have some fun.

So, welcome to ROC. There's more to come...