In recent months, I have spent a great deal of my time talking to governments about science and technology policy and innovation. Money, or more accurately, the lack thereof, is a common theme that runs through all of the conversations, whether in the United States, the European Union, Japan or other parts of Asia. The global economy is still struggling to recover from the recession of 2008, the economic malaise in Europe is very real, and political and economic gridlock in the U.S. threatens the country's ability to chart a competitive future. Despite these challenges, indeed, often because of them, most world leaders are seeking innovative ways to stimulate economic growth and address deep societal problems – aging populations in developed countries, regional environmental concerns and global climate change, shifting education and workforce needs, and rising health care costs.
Today's world leaders and government ministers are surrounded by challenges, buffeted by conflicting (and sometimes irreconcilable) demands, and struggling to cope with rapid technological and societal change. The story is not much better if you are a U.S. governor or the mayor of a large city. Almost all of them all who are willing to be candid will say that the job has never been more difficult. In short, being a senior government leader, at whatever level, is not a great gig right now.
How, you might ask, does this relate to science and engineering research in general, and to high-performance computing in particular? It all devolves to questions of size and scale, something those of us in science understand well, and to a few lessons in realpolitick, where we are often laggards in the classroom of political deal making. (See Being Bilingual: Speaking Technology and Policy)
The Ask and the Close
When most of us are asked to justify government investment in basic research, or to make a case that the government should fund a specific project or research infrastructure, we reach for hoary adages. We claim that we are pushing back the frontiers of human knowledge and seeking answers to questions that have bedeviled or vexed humankind since we first had the intellectual capacity to look around and ask "why?" If truly pressed, we will talk about technology transfer and how research ideas have often birthed multi-billion dollar industries. We may even talk about educating a new generation of scientists and the importance of investment in human capital.
These statements are all true, and like many of you, I have used them many times. They are powerful and effective arguments, given the right initial conditions. However, arguments for science and engineering research funding are not made in vitro; they live or die in vivo. If there were a surfeit of funding, goodness arguments alone can actually suffice. However, I can think of no place, with the possible exception of China, where today's governments find themselves with the budgetary largess to invest in new things. In most cases, governments are struggling to cut or balance budgets, research included.
This brings us to the issue of scale. When initiative or project budgets are measured in the millions of dollars, euros or the equivalent, funding decisions often can be made without rising to the highest levels of political discussion. In most cases, it is a committee or ministerial decision to "plus up" a budget to support such activities. Supercomputing funding was once decided at this level. After all, the U.S. NSF supercomputing centers programs started with an aggregate budget of ~$70M/year in the early 1980s. With single petascale machines costing hundreds of millions of dollars, this is true no longer.
When proposed science and engineering projects or initiatives have estimated costs in the billions of dollars, they cross a critical social and governmental decision threshold. At this level, they compete visibly and directly with other national priorities – national defense, social services, health care, and public infrastructure – and the old arguments are no longer sufficient. A different dialog ensues, one involving participants who are often neither well versed in science and engineering, nor particularly supportive of its interests.
Instead, at the larger scale, one must make arguments for science and engineering investment based on both scientific benefit and other national priorities -- national security (e.g., nuclear weapon stockpile stewardship), national competitiveness (e.g., sustaining a national industrial base) and societal benefit (e.g., creating higher quality, lower cost health care). History has shown that this is difficult, but quite possible. One need only look at the termination of the Superconducting Super Collider (SSC), the fractious debate over completion of the James Webb Space Telescope, and the concerns over continued support for the International Thermonuclear Experimental Reactor (ITER) to see the challenges that accrue to nation-scale and planetary science and engineering projects, particularly when there are cost overruns.
The Exascale Moral
All of our project cost estimates for an exascale initiative place it well in excess of one billion dollars. Thus, the initiative is squarely in the realm of national priority debates, and the science and engineering case, though absolutely necessary, is no longer sufficient. One must also marshal arguments that address national, regional and global needs, and demonstrate differential advantage relative to other competing priorities. This can be done. Remember, the U.S. Accelerated Strategic Computing Initiative (ASCI) was funded using just such a combination of arguments.
Finally, one must create advantage for those government leaders who might be advocates or allies, as well as counter the arguments of opponents. After all, both advocates and opponents face difficult choices and other constituencies who will feel disenfranchised, regardless of the choices. They need arguments that have political heft and popular credibility. This is the essence of realpolitik.
Millions and billions: the difference is not just quantitative; it is politically qualitative. We must couple scientific advances and societal advantage in our arguments to make the case for continued support for scientific computing and exascale infrastructure – hardware, software, and data analysis. We can do this, working together.