Enterprises Embrace Obama’s HPC Moon Shot
First man on moon walking on the moon royalty free stock photograph
Could President Barack Obama's executive order establishing the National Strategic Computing Initiative (NSCI) do for high performance computing what President John F. Kennedy's proclamation about landing on the moon do for the space industry? While the stakes may (literally) not be as high, many enterprises and members of the HPC community are optimistic that the order and the NSCI will spur similar excitement and enthusiasm, ultimately resulting in more widespread adoption of corporate HPC implementations for expanded functions across new departments. Others, however, question the government's role in HPC innovation.
"I believe, like many other HPC professionals and organizations in this arena, that this is indeed a very important statement. The last time such a pronouncement was made was several years ago, during the early days of the George Bush presidency and has languished since then. This helps re-affirm that HPC is a critical technology at a national level and should be treated as such going forward," said Sharan Kalwani, a long-time HPC industry professional.
Unveiled last week, the order's goal is to build supercomputers capable of one exaflop – about 30 times faster than today's speediest machines, according to a statement by Tom Kalil, deputy director for Technology and Innovation at the White House Office of Science and Technology Policy, and Jason Miller, deputy assistant to the President and deputy director of the National Economic Council. Additionally, large-scale computing must address the exabyte storage demands of a big-data world, they wrote. These capabilities would allow businesses to leverage Internet of Things (IoT) in areas such as artificial intelligence (AI), machine and deep learning, sophisticated numerical processing, modeling and simulation, big data, and simulations, they said.
"By strategically investing now, we can prepare for increasing computing demands and emerging technological challenges, building the foundation for sustained US leadership for decades to come, while also expanding the role of high-performance computing to address the pressing challenges faced across many sectors," according to the statement.
If it gets the necessary, ongoing support, Obama's announcement could do for HPC what Kennedy did when he pressed the United States to beat Russia to the moon, suggested Michael Couvillion, chief technology officer at DrillingInfo, in an interview. But to truly be successful, enterprise-wide HPC adoption depends on resolving business needs, he said.
"[The order] starts taking it out of the realm of just the lab guys doing high end lab stuff and puts it in the realm of solving business problems," said Couvillion. "HPC is coming out of the realm of R&D. It's getting to the point where you can have real world impact to real world problems. That's what drives private industry."
These accomplishments also would place the United States back on top of the world's high performance computer pyramid. Tianhe-2, a supercomputer developed by China’s National University of Defense Technology has held the top ranking on the Supercomputer 500 for 2.5 years, edging out Titan, a Cray XK7 system at the Department of Energy’s Oak Ridge National Laboratory. With 233 supercomputers, the US has more systems than any other nation, but that's down from 265 in 2013 and continues a downward trend, the listing reported.
"We don't have the number one supercomputer any more. Another country has that," said Rhonda Chicone, professor at Kaplan University and former chief technology officer at Notify Technology for more than 14 years, in an interview. "We want to maintain our reputation of being able to innovate."
Bragging rights alone may be good for the populace but won't benefit corporate America, said Rasheen Whidbee, systems and network administrator at Vencore. "I do think reclaiming the top spot from China would provide a little boost to the American morale, especially with a change of guards for the presidency, but we really need to address education and how we value science, technology, and exploration," he told EnterpriseTech.
It's exactly for these reasons organizations must get access to high performance computers sooner than later, said Zhong Wang, a computational biologist and big data scientist at the Joint Genome Institute.
"It is obvious to me, a computational biologist, that the data intensive problems we have now will get much worse, datasets of petabyte scale will become a norm. Also the heterogeneity of data will get worse, for example, in the genomics field the need for integrating genomics, transcriptomics, metablomics data will grow significantly," he told EnterpriseTech. "In the short term, this supercomputer will bring immediate benefit to big data mining algorithms, especially to deep neural network based algorithms. In long term, Exascale may enable scientific questions such as modeling a very complex environmental microbiome system using dynamic data collected over a long period of time, and predict its changes in response to environmental changes."
Market Readiness
Enterprise IT departments that must deliver solutions capable of supporting their organizations' analytics, security, research and development, or other compute-intensive applications may be more likely to consider HPC after the president's executive order, said Chicone. As they face increased pressure to do more with less, in real time with even more data, more businesses will come to see HPC as a solution, she said.
"It has to be an incremental thing and we're seeing companies are doing that now, slowly but surely. It's going to depend on the enterprise and what they need to achieve their business goals. I think the executive order is maybe going to get us there but we'll have to take it in chunks and each company is going to have to look at their needs and go from there," she said. "If you have high performance computing and computers with huge amounts of storage, you need systems that take advantage of that high performance computing. Those systems need to be built. Some of the tools we have today don't take advantage of HPC."
Gabriel Broner, SGI
HPC vendors already see steadily growing demand for HPC systems outside traditional segments such as research and development and engineering. Obama's order will, they expect, only hasten demand, R&D, and interest from corporate clients.
"Most of our business is in the enterprise side," said Gabriel Broner, vice president and general manager, high performance computing, at SGI, in an interview. "[HPC] allows them to innovate more freely. HPC is moving more and more into the heart of the enterprise. Today we're seeing some of the more bold CIOs realize some of the benefits of innovating around IT and using in-memory computing. We're seeing that happening today and that's a good indicator that more people will do it in the future. Once the early adopters show good results of using high-performance compute technologies in their environments, more and more in the enterprise will figure it out."
Added Jim Ganthier, vice president and general manager of engineered solutions and cloud at Dell: "The president has validated a lot of what we were thinking. Let's just say Dell and Intel were smiling."
They won't be alone: Most executives predict an influx of new HPC players plus expanded research by today's industry leaders. These developers will create new tools, products, and services for the anticipated market, executives said.
"First, the HPC initiative must go through the research and development phase. Imagine having to purchase a single, highly redundant and modular server at each NOC location instead of a cluster of servers to do the same work. This is what the future of infrastructure looks like. Hardware always develops faster than software and a modular and scalable HPC platform can essentially ensure that enterprise technology corporations are set with low ongoing expansion and maintenance cost," Nick Espinosa, CIO of BSSi2, told EnterpriseTech. "While there will be some pain in the beginning – namely the cost of early adoption – ultimately those who embrace HPC will be in a better position to expand their client and services base for less cost. The next challenge, after HPC comes to fruition, will be to ensure that there is enough fiber run to actually accommodate the bandwidth needs of HPC as well as the public's ever growing demand for more speed and access."
This order encourages small and large, new and existing developers to consider HPC as a viable market for their research and development investments, said Kalwani. "Certainly this will give an incentive to both new developers and existing providers – small and large – to throw their lot towards developing new areas of technology. If anything, this initiative should itself be disseminated much wider, to audiences beyond the traditional HPC, so fresh ideas can be brought in," he said. "In the short term, I expect to see a renewed energy amongst the established players to be a part of the solution. Long term, it should help nurture and grow the critical areas of expertise shortage, newer technologies creation, manufacturing prowess in supporting infrastructures needed for building such machines, as well as a determination in academia to produce skills which can harness these in innovative ways and not the usual incremental extension(s) of older methods."
Enterprise Challenges
However, not all executives agree the presidential initiative will encourage HPC adoption.
"HPC solutions have many hurdles, one being pricing and training. It needs to be within the reach of medium size organizations for the technology to really take off," said Whidbee. "I don't believe this project will have much of an affect on private enterprises, besides the ones that interact with the government on contracts."
Recruiting and retaining qualified technologists already is challenging, agreed Kaplan's Chicone. Finding HPC experts, especially those able to integrate HPC into enterprises' existing infrastructure, governance, and security solutions, will be extremely difficult for all but the most desirable employers, she added.
Today, these computers are too big for some corporate uses, Reuben Yonatan, founder of GetVoIP, told EnterpriseTech.
To succeed in the enterprise, HPC devices must fit in our hands, says Reuben Yonatan.
"To understand how businesses might use supercomputers in the future, look no further than how they are using them now. IBM's Jeopardy-playing supercomputer Watson is being adopted in both for-profit and not-for-profit," he said. "The real question is size. Watson was once the size of a bedroom, now it can fit in a housing the size of about three pizza boxes. Any phone that fits in your pocket would leave a 1990's supercomputer in the dust. But even a pizza-box-sized computer is too big by today's standards for the average user. So I believe that the recently launched initiative won't affect businesses either way until at least one generation after, when the [high performance] computers get small enough to be used by mere mortals. It's a 15-year plan, and it may take 15 years before today's supercomputers can fit in your pocket."
Others question government's involvement in trying to direct technological advancement. The results from Silicon Valley and similar regions around the world typically ignite from entrepreneurs, not government mandates, they said.
"While it is laudable that this president recognizes technology is imperative to compete globally, it is laughable to believe it can be regulated with the motto 'to out compute is to out compete.' Adding a layer of government partnerships and suggestions based on their case studies from government-funded entities would not speed up cloud and technology adoption; the net impact would be the reverse," Mike Torto, CEO of Embotics, told EnterpriseTech. "Businesses innovate best when left to compete via market forces with no political interference or shackles to any one interest."
Although optimistic about the potential, there's concern about the timing and remaining questions.
"The White House’s recent initiative is meant to encourage progress in
HPC and that is very exciting, but we will have to wait until at least 2016 for an agency to draft an official budget and specific number that the government will invest," said Ari Zolden, CEO of Quantum Media Group.
The details are critical, agreed DrillingInfo's Couvillion.
"I am hopeful for the details that will come out because that's where it really will tell the tale," he said. "Without this there's no hope it would progress from where it is today. [HPC has] been around for 20, 25 years and hasn't been mainstreamed yet."
