A new report calls into question the National Science Foundation’s ability to continue its leadership position as a provider of advanced computing resources.
At the behest of the National Science Foundation, a National Research Council committee carried out a study examining the priorities and associated trade-offs for advanced computing in support of NSF-sponsored science and engineering research. The resulting interim report – Future Directions for NSF Advanced Computing Infrastructure to Support U.S. Science and Engineering in 2017-2020 – takes a candid look at an agency struggling to keep pace with growing demand given the high procurement costs and shrinking NSF resources.
“Since the advent of its supercomputing centers, NSF has provided its researchers with state-of-the-art computing systems,” the report notes. “But it is unclear, given their likely cost, whether NSF will be able to invest in future highest-tier systems in the same class as those being pursued by the Department of Energy, Department of Defense, and other federal mission agencies and overseas.”
Multiple trends are converging to make this a trying time for HPC providers. Negotiating the exascale divide will take significant attention to power consumption, inter-chip communications and fault tolerance. To prevent stagnation as CMOS scaling winds down, the community is forced to pursue new system architectures that maximize performance per watt, leaving it to programmers to adapt system software and algorithms accordingly.
There is also more interest in data-intensive computing, which is emerging as a “fourth paradigm” for scientific discovery, complementing theory, experiment, and simulation. “Data-intensive research may require high-performance input/output (I/O) systems, access to very large storage systems using systems with different architectures than traditional high performance computing systems, and new approaches to data visualization,” according to the authors.
Keeping up with demand
The days of free-lunch, Moore’s law subsidized performance gains are nearly over, but demand is higher than ever. “Demand for advanced computing has been growing for all types and capabilities of systems, from large numbers of single-commodity nodes to jobs requiring thousands of cores; for systems with fast interconnects; for systems with excellent data handling and management; and for an increasingly diverse set of applications that includes data analytics as well as modeling and simulation,” observes the committee.
Anecdotal reports suggest researchers are finding it harder to get compute time due to the “double jeopardy” of having two hurdles to clear: obtaining funding for their research proposals and getting the necessary computing allocation. “Given the modest acceptance rates of both processes, such a process necessarily diminishes the chances that a researcher with a good idea can in fact carry out the proposed work,” concludes the committee.
A tipping point
The NSF’s leadership standing has fallen behind that of other mission-oriented agencies in the United States, such as the Department of Energy, and international research organizations, such as the Partnership for Advanced Computing in Europe or the Ministry of Science and Technology in China. These groups are pursuing systems at least an order of magnitude more powerful, for both computation and data handling, than current NSF systems.
The committee describes an already strained situation that will be further challenged by an influx in research data, growing demand for resources and the demise of Moore’s law.
“It is becoming increasingly difficult to balance investments in advanced computing facilities, given the large and growing aggregate demand, the steep cost of the highest-end systems, growing demand for data-intensive as well as compute-intensive systems, and the constant or shrinking NSF resources,” write the authors. “Compounding the challenge is the wide variety of computing needs, the state of scientific data and software, and wide variation in ability to effectively use advanced computing across scientific disciplines. Moreover, the range of science and engineering research sponsored by NSF involves a diverse set of workflows, including those that involve primarily compute- or data-intensive processing and ones that involve combinations of both.”
Alternative “high-tier” options offered up for consideration include purchasing computing services from other federal agencies or making arrangements with commercial services. While the committee recognizes that they are not a panacea for all workloads, cloud computing and publicly available but privately held data repositories could open the door to large-scale compute and data capabilities to underserved researchers and fields.
The report also urged greater focus on advanced computing within the NSF strategic plan given the essential role that HPC plays in modern research.
The final report, which can be downloaded here, will include a framework for future decision making regarding how to balance competing demands for cyberinfrastructure investments. The committee will not make specific funding recommendations, but they are mindful of the need for timely input for the NSF budget process. The science and engineering communities that use, develop, and provide advanced computing capabilities are encouraged to give feedback.