The National Science Foundation (NSF) asked the National Research Council to study anticipated priorities and associated trade-offs for advanced computing in support of NSF-sponsored science and engineering research. (See Box P.1 in the Preface for the complete statement of task.) This interim report contains a preliminary set of issues the Committee on Future Directions for NSF Advanced Computing Infrastructure to Support U.S. Science in 2017-2020 believes that NSF, the science and engineering research community, and the committee itself need to consider. It is intended to stimulate discussion and prompt feedback that the committee will consider in preparing its final report. (See the Preface for how to provide feedback to the study committee.)
BUILDING ADVANCED COMPUTING INFRASTRUCTURE TO SUPPORT INTEGRATED DISCOVERY
Advanced computing in this context refers to the technical capabilities that support compute- and data-intensive research across the entire science and engineering spectrum and that are so expensive that they are shared among multiple researchers, institutions, and applications. Compute-intensive modeling and simulation, the historical focus of high-performance computing systems and programs, is an established peer, standing beside theory and experimentation, in the scientific process. Data-intensive computing is emerging as a “fourth paradigm” for scientific discovery, complementing theory, experiment, and simulation, and
may require new technical and programmatic responses. Compute- and data-intensive approaches are increasingly used in combination: data is used to validate models, simulations are used to quantify uncertainty or fill in for incomplete theory, and stochastic models link modeling and data analytics. Data-intensive computing is becoming more important as the volume of data grows, as new analytical techniques are adopted, and as some fields move from being primarily compute-intensive to being much more data-intensive.
For its final report, the committee will explore and seeks comment on
1. How to create advanced computing infrastructure that enables integrated discovery involving experiments, observations, analysis, theory, and simulation.
TECHNOLOGY CHALLENGES
Unfavorable trends in power consumption and inter-chip communications are forcing consideration of new system architectures, the development of new algorithms and software approaches to use them, and more attention to redundancy and fault tolerance. Absent new technology, the anticipated end of sustained reductions in the ratio of price to performance (a benefit of Moore’s Law) portends stagnation in computer performance improvement. For data-intensive systems, variability in storage hardware performance and failure rates constrain the performance and practical size of very-large-scale systems. Also, it will not be straightforward in all cases to keep scaling up system and scientific software to meet growing needs. The resulting uncertainty about technical direction complicates planning for future extreme-performance computers.
Today’s approach of federating distributed compute- and data-intensive resources to meet the increasing demand for combined computing and data capabilities is technically challenging and expensive. New approaches that co-locate computational and data resources might reduce costs and improve performance. Recent advances in cloud data center design may provide a viable integrated solution for a significant fraction of (but not all) data- and compute-intensive and combined workloads.
New algorithms and software approaches will be needed to effectively use systems with new architectures, and they can also play an important role in continuing to improve the performance of scientific codes and the productivity of researchers. Some developments may best take place within individual research areas and disciplines, but others may benefit from common, coordinated efforts.
New knowledge and skills will be needed to effectively use these new advanced computing technologies. “Hybrid” disciplines such as computational science and data science and interdisciplinary teams may come to play an increasingly important role.
For its final report, the committee will explore and seeks comments on
2. Technical challenges to building future, more capable advanced computing systems and how NSF might best respond to them.
RESPONDING TO GROWING DEMAND
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.
Anecdotal reports point to a low and perhaps declining rate of success for obtaining allocation of time on existing machines. Given the “double jeopardy” that arises when researchers must clear two hurdles—first, to obtain funding for their research proposal and, second, to be allocated the necessary computing resources—the chances that a researcher with a good idea can carry out the proposed work under such conditions is diminished.
Since the advent of its supercomputing centers, NSF has provided its researchers with state-of-the-art computing systems. 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. Options for providing highest-tier capabilities that merit further exploration include purchasing computing services from federal agencies (thus increasing access beyond that driven by direct mission interests) or by making arrangements with commercial services (rather than more expensive purchases by individual researchers).
More broadly, across a wide spectrum of system capability, the growth of new models of computing, including cloud computing and publically available but privately held data repositories, opens up new possibilities for NSF. Access to these commercial facilities could widen access to large-scale capabilities for computation and data analytics, but the cost trade-offs are complicated and need to be looked at carefully.
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. 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.
It is thus harder than ever to understand the expanding and diverse requirements of the science and engineering community; explain the importance of a new, broader range of advanced computing infrastructure to stakeholders, including those that set budgets; explore non-traditional approaches; and manage the advanced computing portfolio strategically.
3. The committee will review data from NSF and the advanced computing programs it supports and seeks input, especially quantitative data, on the computing needs of individual research areas.
For its final report, the committee seeks comment on
4. The match between resources and demand for the full spectrum of systems, for both compute- and data-intensive applications, and the impacts on the research community if NSF can no longer provide state-of-the-art computing for its research community.
5. The role that private industry and other federal agencies can play in providing advanced computing infrastructure—including the opportunities, costs, issues, and service models, as well as balancing the different costs and making trade-offs in accessibly (e.g., guaranteeing on-demand access is more costly than providing best-effort access).
6. The challenges facing researchers in obtaining allocations of computing resources and suggestions for improving the allocation and review processes for making advanced computing resources available to the research community.
POSSIBLE NSF RESPONSES
Better Understanding of Science and Engineering Opportunities, Priorities, and Requirements for Advanced Computing
Not all research areas or programs have defined their requirements for advanced computing or established processes for regularly updating and refining them, such as by constructing roadmaps that describe science or engineering goals and advanced computing resources needed.
Such analyses may provide useful information for understanding aggregate capability and capacity needs and expected trends in these needs, for understanding overall NSF resource requirements, for prioritizing investments, and for better aligning research programs and supporting advanced computing investments. Because scientists can effectively use infrastructure only when it is presented as an integrated whole—encompassing appropriate hardware, software, data, networking, technical services, and so forth; it may be most productive to use a functional rather than a technology-focused or structural approach focused on individual elements.
For its final report, the committee will explore and seeks comment on
7. Whether wider collection and more frequent updating of requirements for advanced computing could be used to inform strategic planning, priority setting, and resource allocation; how these requirements might be used; and how they might best be developed, collected, aggregated, and analyzed.
Enhanced Organizational Stability and Flexibility of NSF-Funded Advanced Computing Centers
Although NSF’s use of frequent open competitions has stimulated intellectual competition and increased NSF’s financial leverage, it has also impeded collaboration among frequent competitors, made it more difficult to recruit and retain talented staff, and inhibited longer-term planning.
For its final report, the committee seeks comment on
8. The tension between the benefits of competition and the need for continuity as well as alternative models that might more clearly delineate the distinction between performance review and accountability and organizational continuity and service capabilities.
Enhanced Strategic Planning and Internal Coordination
Advanced computing receives less attention in the current NSF strategic plan than might be expected, given its vital role in science and engineering, although it is the subject of a separate strategy focused on cyberinfrastructure. Decision making about advanced computing is distributed across the Division for Advanced Cyberinfrastructure, other divisions and division programs, the Major Research Instrumentation Program, and individual research institutions. Both coordination and
strategic decision making seem especially important in an era of growing demand and cost. Top-down mandates often prove ineffective, even when the coordination is very much needed, and reaching consensus through “grass-roots” efforts may be too slow. Both top-down and bottom-up processes require mechanisms for identifying detailed needs of the directorates and their programs and for ensuring adequate community input.
For its final report, the committee seeks comment on
9. How NSF might best coordinate and set overall strategy for advanced computing-related activities and investments as well as the relative merits of both formal, top-down coordination and enhanced, bottom-up process.
