The National Strategic Computing Initiative, Pt. 2 – The NITRD
Published: August 19, 2015
The NSCI has the potential to make High Performance Computing R&D one of the fastest growing areas of big data related work in the federal government.
Last week’s post on the National Strategic Computing Initiative examined areas of potential business opportunity in High Performance Computing R&D at several federal agencies. One of those areas was only tangentially touched upon – the $242 million FY 2016 Networking and Information Technology Research and Development (NITRD) program budget request – even though the organizations comprising NITRD are central to NSCI-related efforts. In order to square the circle, this week’s post takes a closer look at the NITRD budget request for the coming fiscal year.
The best place to start is probably with a brief description of the NITRD program. NITRD is a consortium of 20+ federal agencies under the coordination of the White House’s Office of Science and Technology Policy (OSTP) which “coordinates federal R&D investments in advanced digital technologies that are essential to the economic growth and prosperity” of the United States. NITRD agencies coordinate the planning and execution of their research programs “to leverage investments in ways that have a greater positive impact than agencies could achieve [by] working alone.”
NITRD investments are divided into 8 interest areas, of which High Performance Computing, or High End Computing (HEC), as NITRD puts it, is only one. The other areas deal with cyber security, big data, software development, large scale networking, and so on. Essentially, NITRD is the government’s point organization for researching emerging technologies and their application by federal agencies.
Every year the NITRD program submits a supplement to the president’s budget that is highlighted in the Analytical Perspectives volume of the request. This supplement provides insight into where the member agencies are either spending or planning to spend on technologies in each of NITRD’s 8 interest areas. Concerning HEC in particular, the table below summarizes investments in two areas of High End Computing that are outlined in the NITRD supplement: infrastructure and Applications and R&D.
The number of dollars dedicated to HEC has been rising steadily over the last few years, with investment in R&D leading the way in percentage terms. Based on these numbers we can conclude that federal agencies spend an average of $1.5 billion per year on HPC; and if the NSCI gains momentum we can expect the R&D portion of that spending to be where most of the new money will be found.
HEC R&D Budgets by Agency
Breaking down the request by agency, allows us to hone in on the places where HEC spending on R&D will be the highest (see table below). The agencies where an increase in R&D budget dollars is expected are highlighted in dark green, while those in red are agencies where the total requested dollars are falling.
The increases expected at DOE and NSF are the largest for reasons explained in the previous post on the NSCI. DoD shows a net decrease across all organizations that work with HPC, so even though the NSCI stipulates that DoD will focus primarily on developing data analytics, the overall amount of money dedicated to that work is declining. It is nevertheless encouraging to see that the DoD budget for HEC R&D remains over $200 million.
HEC R&D Priorities at DoD, DOE, NSF, and DARPA
Space will not allow an extensive discussion on the types of HEC R&D work that will be done at the four agencies with the highest FY 2016 budget requests, so I’ll stick to summarizing their priorities. Vendors who do this work will recognize the areas of opportunity related to big data and, potentially, cloud computing. Greater detail on these priorities can be found in the FY 2016 NITRD Budget Supplement.
DoD (Ex. DARPA)
Application development environments and algorithms, runtime environments, system software for extreme-scale computing, augmenting large-scale HPC systems with targeted special-purpose or emerging computational platforms for niche and future application-space requirements, developing scalable, complex, multi-physics-based codes for critical defense applications, developing advanced cybersecurity tools.
Research to support data-intensive science for exascale computing, machine learning for adaptive systems and analytics, cybersecurity, tools for software development and system design, designing user interfaces, visualization and analytics for understanding extreme scale datasets.
Operating/runtime systems, development environments, productivity tools, languages, compilers, libraries (eXploiting Parallelism and Scalability [XPS]), quantum and biological computing, cyber-physical systems; cybersecurity (Secure and Trustworthy Cyberspace [SaTC]); networking, including integrative data cyberinfrastructure, the NSF Big Data Initiative (DIBBS, RDA, EarthCube).
Research computing technologies critical to enable data intensive embedded systems, including energy efficient technologies and architectures, unconventional computation and data representations based on probabilistic inference, thermal techniques and approaches, photonic-based technologies and architectures, cybersecurity.
Analysis of the NITRD data shows that the overall number of budget dollars dedicated to HPC-related R&D will grow 16.8%, from $524 million received in FY 2014 to $612 million requested in FY 2016, if the FY 2016 budget is passed by Congress and approved by the president. HPC is a very specialized field of work that heretofore has mostly benefitted vendors selling large supercomputers. The renewed focus on R&D prompted by the NSCI, however, promises to change this by funding a large number of smaller HPC-related projects. This potentially will make HPC R&D one of the fastest growing areas of big data related work in the federal government in the years to come.