In 2010, IDC’s EMEA division set forth a number of recommendations for strengthening Europe’s HPC presence in the report, Development of a Supercomputing Strategy in Europe, which included a broad set of action items to propel Europe into the exascale era.
In the report, the research group suggests that in order for Europe to reach the goal of being a center for high performance computing innovation by 2020, a number of challenges must be addressed. IDC listed a number of new investment areas, including expanding the number and size of HPC resources in the EU, creating a set of exascale labs and testbed centers to address software barriers, attracting new students to HPC, and more generally, enhancing leadership in strategic application areas.
One major facet is noticeably absent from this litany of sustained fixes—one that Bull’s Matthew Foxton addressed during a talk at the International Supercomputing Conference (ISC) last month in Germany.
According to Foxton, Europe’s challenges when it comes to the future of high performance computing in Europe and the coming era of exascale go far beyond leadership enhancements and a focus on software. As he put it, there is a great emphasis on R&D in Europe in general, but where Europe falls short is placing too much attention on the “R” and not enough on the “D”.
Foxton says that the great missing link in discussions about the next generation of technology in Europe is the near complete reliance on other regions for the hardware resources. Without that holistic technological approach coming from within (software, research, development and hardware), Foxton says that Europe could be left behind by the very regions that are supplying critical HPC hardware infrastructure.
Your bias bell probably went off at the relationship between Bull and their plea for strengthening European HPC hardware markets. It’s natural that one of the few European vendors of HPC hardware would want to enhance its market share with what amounts to a patriotic call to arms. However, despite this apparent self-interest, there is something to Foxton’s claim. Why isn’t Europe a center for HPC hardware, at least to the extent that it is feeding its homegrown software and research prowess into foreign-made machines?
Foxton pointed to the estimate that nearly 50% of all HPC spending in Europe was consumed by hardware purchases, yet notes that Europe is not a major player. As he said during his talk, “For a market that is growing—one that is a training ground for engineers, a source of employment for these engineers—why are we a world-class buyer of HPC but not a world-class provider?”
He claims that the answer doesn’t lie in a European Buy Act that would force EU-internal purchasing decisions but rather that Europe needs to move beyond its traditional scope in where it focuses HPC development. He points to a question that he says has arisen in conversations about the topic: “Are the Europeans themselves neglecting Europe?”
“We can listen to analysts and commentators when they say we need to just buy a certain exascale system versus just building the first one ourselves. But then, who do we buy from? America? China? Japan? India?… We could buy from any of the world’s major regions but why would we be the only region in the world not capable of producing these systems? Do we really believe encouraging this kind of goal would really encourage the European industry?”
Beyond the roundabout way of encouraging European HPC by looking outside for key hardware investments, Foxton makes another point. He says that this practice of “neglect” is going to start having a visible impact on the most important element of Europe’s HPC future—young, skilled engineers.
He says that among all the threats to European supercomputing and all of the economic incentives it can provide is the risk of brain drain. This is not a new topic for European HPC professionals and has formed the center of a number of working group topics.
For instance, the European Commission has set up a number of working groups to indentify some of key components of Foxton’s argument for a more stable, productive European HPC industry. The ESFRI group works with the e-Infrastructure Reflection Group (e-IRG) to promote the possibilities offered by a more connected HPC ecosystem. The group says that “E-infrastructures stimulate the identification and creation of new scientific communities, uniting researchers who are working on similar challenges and are willing to share resources and reach new levels of collaboration.”
Moreover, they claim that “Widespread use of e-infrastructures represents an effective answer to problems such as the digital divide and brain drain. This is demonstrated by the Large Hadron Collider at CERN, which is serving the worldwide community of particle physicists.”
Even still, projects like the LHC are based on software and academic research—not exactly the broad-reaching economically appealing application areas that Foxton sees as being the most valuable. He pointed to key areas in oil and gas, the aerospace industry, and high energy physics to support lasting clean energy solutions as being crucial to the economic well-being of Europe.
This is where his argument about the missing D in R&D is most keenly felt—if Europe is providing the world with cutting-edge research and applications to be computed elsewhere or on foreign-made hardware, could they be missing out on a large slice of the HPC pie that lies waiting for nothing more than a robust competitive HPC hardware ecosystem?