US adopts Soviet-style five year plans for supercomputing
In a move that echoes past Soviet and current Chinese strategies of pursuing ‘five year plans’ for economic and technology development, President Obama has committed the United States to a five-year National Strategic Computing Initiative (NSCI) to promote high-performance computing (HPC).
In an Executive Order issued on 29 July, the US President presented the NSCI as ‘a whole-of-government effort designed to create a cohesive, multi-agency strategic vision and Federal investment strategy, executed in collaboration with industry and academia, to maximise the benefits of HPC for the United States’.
No new funding has been made available. Instead, the Order establishes an inter-agency Executive Council. This is to be chaired jointly by the President's Science Advisor (currently Dr John Holdren) and the Director of the Office of Management and Budget (currently Shaun Donovan). Formally, Dr Holdren is Director of the Office of Science and Technology Policy (OSTP) within the White House, and he will nominate the other members of the Council.
The US five-year plan for HPC
Over the next three months, the Executive Council is to draw up ‘an implementation plan’ to support and align efforts across Government agencies in support of the NSCI. This will be the US’s national five year plan for supercomputing, as the Executive Council is being required to monitor and update the plan annually over a period of five years ‘and document the progress made in implementing the plan, engaging with the private sector, and taking actions to implement this order’.
If implemented by his successors, President Obama’s initiative will provide for at least three five-year plans. One of its strategic objectives is ‘establishing, over the next 15 years, a viable path forward for future HPC systems even after the limits of current semiconductor technology are reached (the "post-Moore's Law era")’.
The proposal for the NSCI explicitly recognises the convergence of compute-intensive and data-intensive systems, with a background briefing document, NSCI Fact Sheet, describing one of the priorities as being to ‘create systems that can apply exaflops of computing power to exabytes of data’. It is envisaged that US Government agencies will work with computer vendors to create advanced systems for applications involving combinations of modelling, simulation, and data analytics. US Government research programmes will develop new approaches to hardware, system architectures, and programming tools. Moreover, the agencies have been given explicit responsibility to ‘foster the transition of these technologies from research to deployment’.
Response to China?
Interestingly, given that China has had the most powerful supercomputer in the world since June 2013 as measured on the Top500 list, the Fact Sheet uses the past tense when it insists that ‘the United States has been [emphasis added] the leader in building large-scale computing systems and in applying them for modelling and simulation’.
China will soon have not one but two 100 petaflop machines using its own domestically development CPUs or accelerators. In what is perhaps a veiled reference to the Chinese developments, the briefing document recognises that ‘Other countries have undertaken major initiatives to create their own high-performance computer technology’. Therefore it concludes: ‘The Nation [i.e. the USA] must preserve its leadership role in creating HPC technology and using it across a wide range of applications’.
In addition to the drive to develop new hardware over the next 15 years, efforts to keep up with the international competition on Exascale systems, and build up US expertise generally, there are two further strategic themes of the NSCI.
One recognises that current HPC systems are very difficult to program and that shifting a program to a new machine is not straightforward. Work under this thematic priority will address the level of expertise and effort required to develop HPC applications, because they pose a major barrier to more widespread use.
A related issue forms another theme: making HPC readily available. It will take steps to address the fact that many companies and many research projects that could benefit from HPC technology, but lack expertise and access.