The venerable American technology icon Cray has seen its share of changes, market cycles and economic downturns. Through the last 13 years, the firm, led by CEO, President and supercomputing veteran, Pete Ungaro, has continued to invest heavily in R&D to build some of the most powerful and innovative supercomputers on the planet.
I talked to Pete about the convergence of big data and supercomputing, the future of exascale – building supercomputers that perform an incomprehensible billion billion calculations per second and how Cray has retained its relevance in the technology landscape since its founding in 1972.
M.R. Rangaswami: What does the convergence of big data and supercomputing mean for the future of business?
Pete Ungaro: As many of us know, the growth in data is exploding with no end in sight. Instead of just math models that have been the purview of traditional supercomputing modeling and simulation, we are now also using these systems to calculate data models found in areas such as AI (artificial intelligence) and big data analytics. For example, we can now predict the weather and extreme weather events such as hurricanes with amazing accuracy. Scientists are taking these scientific models of the earth’s weather system and incrementing it with data models on both the front and back end (and sometimes even within the simulation itself) to further improve the accuracy and fidelity of the prediction.
The results of these data models feed even more simulations, and we get a cycle that as our data sources grow we can use this new data to improve our predictions. The combination of math models and data models give us the ability to answer tougher, bigger and deeper questions. We see the worlds of traditional supercomputing and high-end analytics and AI converging. In fact, I would say that before too much longer supercomputing will no longer be a standalone, niche market like it has been for so many years.
We will soon be talking about a high-performance data-intensive computing world and as such, supercomputing and the supercomputing market will change and broaden to a far larger market. Already, many business leaders are noting that as the size of their companies’ datasets grows, it becomes too hard for traditional commodity clusters or enterprise-class systems to model this data. One example is our insurance company customer; as it gathers more and more data, it wants to gather more intelligence from that data. But what this customer has found is that they cannot get what they want or need from their traditional enterprise systems or going to typical sets of resources found on the public cloud. That’s where a solution like Cray systems come into play and for us -it’s an exciting expansion of our market.
M.R.: How do you view Cray’s role in the exascale arms race?
Pete: While “exascale arms race” is kind of a catchy term, I don’t think it accurately describes what is – a journey. At Cray, we’re not just marching toward this destination called exascale, and then we stop and proclaim the job done. Exascale is really just a metaphorical mile marker on the journey to expanding the technology and capability of high-end computing systems. Computing at scale is where Cray is at its best and “at scale” never stops. It just keeps scaling.
Also, exascale is really the other side of the big data movement. It’s about figuring out the next generation of supercomputers that will be more than an order of magnitude larger than the fastest supercomputers of today. At Cray, we expect to have the first exascale system online in 2021. That won’t just be the pinnacle, we will be broadening the availability of that technology, making it available in smaller packages so more companies can take advantage of it.
Over time, I expect that many Fortune 500 companies will have an exascale system to help them get insight from their growing data – these powerful systems just won’t be for the national laboratories. In fact, many customers across energy, manufacturing, life sciences and financial services industries have told me that they already have problems that they could use an exascale system to help them solve. This implementation would give them an important advantage over their competition.
As it stands today, Cray has the most systems of any vendor in the Top 100 supercomputers in the world, and we will play an important role in the development of this technology going forward. That is not just for the first exascale systems out there, which may cost hundreds of millions of dollars, but this exascale technology will be used in smaller systems that are vastly more powerful than anything currently available in enterprise-class systems today. These smaller systems may be as small as a single cabinet costing less than a million dollars, broadening their reach to a much larger number of potential businesses that will take advantage of exascale.
For business leaders not following the “exascale arms race,” what they need to know is that there is technology being developed that can solve their toughest business problems. Systems that will give them the competitive edge they are looking for by leveraging their vast amounts of data by using AI, analytics and engineering, and scientific models. As problems that organizations seek to solve need more and more data become more complicated, Cray’s technology becomes more important.
M.R.: As a company that has been around since the 1970s, how is Cray still relevant in today’s technology landscape?
Pete: What makes us relevant today is that the world of data is growing and will continue to. We’re already seeing this in the business world where the problem sizes are getting larger as organizations have more data to compute and derive insight from. This trend is occurring both in industry and in science — one of our traditional markets. Cray’s forthcoming systems will be key to answer these more complex questions and finding better, smarter answers.
We’ve had a consistent focus since Cray was founded: how do we help our customers solve their thorniest and biggest questions. That said, much has changed in the last 40-plus years. Back in the day, we had to develop our own technology from the ground up: processors, interconnects, operating systems and entire software stacks. Now, however, thanks to the evolution of the technology industry at large, we can leverage commodity processors, Linux operating systems, ideas from the world of cloud computing such as containers, to make our technology usable by everyone and not as proprietary (or expensive) as it used to be.
We build each Cray system uniquely for each customer. Where we see gaps in the currently available technology stack, areas where commodity systems can’t get to, Cray supplements this with our own technological and design expertise in the areas of compute, storage and software. In today’s data-driven age, we see a convergence of the high-end big data analytics world, AI, modeling and simulation. That’s what we’re targeting today as a company, and I think it’s going to be a fast growing market.
We’re still sticking to our roots but we’re more focused than ever on expanding our capabilities to be usable for a far broader range of customers.
M.R. Rangaswami is the co-founder of Sand Hill Group and publisher of SandHill.com.