nsf.gov - S&E Indicators 2014 - Chapter 5. Academic Research and Development - Infrastructure for Academic R&D

Chapter 5. Academic Research and Development

Infrastructure for Academic R&D

Physical infrastructure is an essential resource for the conduct of R&D. Not long ago, the capital infrastructure for R&D consisted primarily of research space (e.g., laboratories and computer rooms) and instrumentation. Accordingly, the square footage of a designated research space and counts of instruments have been the principal indicators of the status of research infrastructure.

Advances in information technology have brought significant changes to both the methods of scientific research and the infrastructure necessary to conduct R&D. The technologies, human interfaces, and associated processing capabilities resulting from these innovations are often called cyberinfrastructure.

Cyberinfrastructure has become an essential resource for science. It helps researchers process, transfer, manage, and store large quantities of data. Cyberinfrastructure includes resources such as high-capacity networks, which are used to transfer information, and data storage systems, which are used for short-term access or long-term curation. It may also involve HPC systems used to analyze data, create visualization environments, or facilitate remote use of scientific instrumentation (NSF 2012). Indicators for research facilities, research equipment, and cyberinfrastructure are highlighted below.

Research Facilities

Research Space

The nation’s research-performing colleges and universities had 202.9 million net assignable square feet (NASF) of research space available at the end of FY 2011 (appendix table 5-8).^[16] This was 3.5% above the net assignable square footage at the end of FY 2009 and continued more than two decades of expansion. However, this increase was less than the median growth (4.7%) for all biennial periods measured from FY 1988 to FY 2011 (figure 5-9).

Biological and biomedical sciences continued to account for the bulk of growth, increasing by 8.0% during the FY 2009–11 period (appendix table 5-8). This field accounted for the largest portion of research space (26.8%), which totaled 54.3 million NASF.^[17] From FY 2001 to FY 2011, research space in biological and biomedical sciences grew 64.5% (figure 5-10). The related field of health and clinical sciences was the second largest in FY 2011, accounting for 36.7 million NASF and 18.1% of the total. Still sizable are engineering (31.7 million NASF, 15.6%); physical sciences (29.6 million NASF, 14.6%); and agricultural and natural resources (27.6 million NASF, 13.6%). Excluding biological and biomedical sciences, total S&E research space has grown only 1.4% since FY 2005. The growth rates have varied across the S&E fields (appendix table 5-8). The computer and information sciences, engineering, and psychology have all increased research space by at least 10%, while space devoted to the other broad science fields has declined or remained the same.^[18]

New Construction

New research space is added each year through new construction projects and the repurposing of existing space. Along similar lines, some space is withdrawn from use. The net result has been an increase in research space for more than two decades. As part of this process, academic institutions broke ground on 8.1 million NASF of new S&E research space construction projects in FYs 2010–11. This total is 50% lower than NASF constructed in FYs 2002–03 (table 5-7). Although the growth rate of new construction projects has declined over the past decade, institutions initiated new construction in all fields in this latest period. The health and clinical sciences and the biological and biomedical sciences fields both saw 2.0 million NASF or more of new construction initiated. Engineering research space construction accounted for 1.3 million NASF. No other fields added more than 0.9 million NASF through new construction during this time.

Academic institutions draw on various sources to fund their capital projects, including the institutions’ own funds, state or local governments, and the federal government (appendix table 5-9). Institutions provide the majority of funds for construction of new research space, typically accounting for over 60.0% of the cost. For the construction of new research space initiated in FYs 2010–11, 61.9% of the funding came from institutions’ internal sources, 30.5% from state and local governments, and the remaining 7.6% from the federal government. The percentage of this funding from institutional sources has remained the same since FYs 2006–07.^[19] The federal portion of funding has been under 10.0% in recent years but declined to 3.2% in FYs 2008–09 before this recent bounce.

Repair and Renovation

Academic institutions expended $3.5 billion on major repairs and renovations of S&E research space in FYs 2010–11 (appendix table 5-10).^[20] They anticipated $3.1 billion in costs for planned repair and renovation of research space with start dates in FYs 2012–13. Nearly $1.0 billion was planned to improve space in biological and biomedical sciences as well as close to $1.0 billion for improvements to health and clinical sciences space. In addition to these slated improvements, academic institutions reported $4.8 billion in repair and renovation projects from their institutional plans that were not yet funded or scheduled to start in FYs 2012–13. An additional $2.6 billion in needed improvements were identified that lay beyond institutional plans. The total backlog of deferred improvements was greater than all projects started or planned for the FY 2010–13 period. The costs for deferred repairs and renovations have consistently been greater than those started or planned for similar cycles in the past.

Research Equipment

In FY 2012, about $2.0 billion in current funds were spent for movable S&E academic research equipment necessary for the conduct of organized research projects (appendix table 5-11).^[21] This spending accounted for 3.2% of the $62.3 billion of total academic S&E R&D expenditures. Spending decreased 11.6% from FY 2011 to FY 2012 when adjusted for inflation. Expenditures for academic research equipment reached the highest mark in several decades in FY 2004. Due in part to ARRA funding, research equipment expenditures approached this level again in FYs 2010–11. After this temporary increase, the FY 2012 expenditures fell to the lowest level measured in constant dollars since FY 2001.

Research equipment expenditures continue to be concentrated in just a few S&E fields. In FY 2012, three fields accounted for 85.8% of the annual total: life sciences (41.0%), engineering (28.1%), and physical sciences (16.7%). The shares for these three fields have remained similarly predominant for many years (appendix table 5-11). Even so, when adjusted for inflation, the annual level of equipment spending in engineering, physical sciences, and the largest life sciences subfields of biological sciences and medical sciences declined from FY 2011 to FY 2012 to pre–FY 2010 levels (figure 5-11).

Some academic research equipment funding comes from the federal government. These federal funds are generally received as part of research grants or as separate equipment grants. In FY 2012, the federal government supported 57.0% of total academic S&E research equipment funding, which marked a 6 percentage point decline from the 25-year high reached in FY 2011 (appendix table 5-12). The federal share of funding varies significantly by S&E field, ranging from 34% to 84% in FY 2012. Atmospheric sciences had the largest proportion of federally funded R&D equipment (83.6%), with astronomy (83.4%) and physics (80.8%) ranking just behind. Agricultural sciences (34.1%) received the smallest share of federal research equipment funding, followed by civil engineering (37.2%).

Cyberinfrastructure

Academic institutions continue to enhance their cyberinfrastructure, which is an essential component to both research and instruction. The cyberinfrastructure indicators noted here include access to high-speed/high-capacity bandwidth, dark fiber, HPC, and the ability to store large amounts of data for immediate access or long-term curation.

Networking

Networking is an essential component of cyberinfrastructure. It facilitates research-related activities such as communication, data transfer, HPC, and remote use of instrumentation.^[22] Universities may have networks that are available to the entire campus community for both research and nonresearch activities. The traffic on these campus networks cannot be differentiated between administrative, instructional, research, and general student purposes. Thus, total bandwidth capacity cannot be treated as an indicator solely of research capacity, and changes in research uses cannot be inferred from changes in bandwidth capacity.

Some cyberinfrastructure is dedicated primarily to research activities. For example, research-performing universities may have access to high-performance networks such as Internet2, an organization established in 1997 that is composed of research, academic, industry and government partners, and National LambdaRail, a university-owned organization established in 2003 that manages a 12,000-mile high-speed network.^[23] The Energy Sciences Network, a DOE-funded network supporting 30 major DOE sites as well as researchers at universities and other research institutions, serves a similar purpose. Regional networks or gigapops (gigabit points of presence) facilitate access by providing networking resources and supplemental bandwidth to the national networks, which are often referred to as the network backbone. These resources are provided to universities as well as government agencies, federally funded research and development centers (FFRDCs), and other entities. The regional networks not only serve as network access points, they also provide advanced network services to ensure reliable and efficient data transfer.

By FY 2012, access to high-performance networks had become widespread at research universities, which is evidenced by the 63% of institutions reporting bandwidth of at least 1 gigabit per second (Gbps) (table 5-8). Thirty percent of academic institutions anticipated network connections of 10 Gbps or greater in FY 2012, compared with 15% of institutions with such access in 2009.

Doctorate-granting institutions have significantly higher bandwidth capacity than non-doctorate-granting institutions due to their research demands. In FY 2011, the percentage of doctorate-granting institutions with bandwidth of at least 2.5 Gbps (43%) was more than 10 times greater than that of non-doctorate-granting institutions (4%). Furthermore, in FY 2012, 53% of doctorate-granting institutions estimated that they would have bandwidth of 2.5 Gbps or greater, compared to 5% of non-doctorate-granting institutions.

Dark fiber is fiber-optic cable that has already been laid but is not yet being used. The amount of dark fiber controlled by institutions indicates the ability to expand existing network capabilities, either between existing campus buildings or from the campus to an external network. The percentage of academic institutions with these unused cables has increased steadily in recent years. The percentage of institutions with dark fiber to their Internet service provider has grown from 29% in FY 2005 to 47% in FY 2011. The percentage of institutions with dark fiber between their own buildings remained high throughout this period, increasing slightly from 86% in FY 2005 to 90% in FY 2011.

High-Performance Computing

Many academic research institutions manage their HPC resources through a distinct organizational unit within the institution that has a separate staff and budget. A total of 192 academic institutions reported ownership of centrally administered HPC resources in FY 2011.^[24] This approach enables faculty to focus on their primary responsibilities instead of being diverted by administration and fundraising to support their own HPC. Central HPC administration can decrease overall operating expenses and create wider availability of computing resources.^[25] However, many HPC resources, not included here, reside beyond direct institutional administration because they are supported by external funding sources.

Forty-seven percent of doctorate-granting institutions provided centrally administered HPC resources, compared to less than 9% of non-doctorate-granting institutions. Similar percentages of public doctorate-granting (48%) and private doctorate-granting (45%) institutions provided these resources. Clusters are the most common centrally administered HPC architecture used by academic institutions because they provide the most flexibility and cost efficiency for scaling in addition to their generally lower administrative costs. Over 97% of HPC-providing institutions employ cluster architectures (appendix table 5-13). HPC-providing institutions also use architectures such as massively parallel processors (11% of institutions), symmetric multiprocessors (19%), or other types of architectures (20%), all of which can be used in conjunction with or as an alternative to clusters.^[26]

Colleges and universities often share their HPC resources with external organizations. In FY 2011, these partnerships most often involved other colleges or universities (72%). Sharing of HPC resources with other external users was fairly evenly distributed among government (21%), industry (18%), and nonprofit organizational (17%) partners. Public institutions were more likely to have external users of their HPC than were private institutions.

Data Storage

As the collection of massive data sets has increased in recent years, data storage and curation have become an increasingly critical issue. Data management plans are often required in funding proposals where large data sets will be used. Of the academic institutions with centrally administered HPC in FY 2011, 56% reported usable online storage greater than 100 terabytes.^[27] A smaller share of public (21%) and private institutions (18%) provided greater than 500 terabytes of online storage.

As of FY 2011, 45% of institutions with centrally administered HPC reported no archival storage. Archival storage includes online and offline storage for files and data that do not support immediate access from HPC resources. This percentage changed little from FY 2009 (43%), yet it stands much higher than FY 2007 (29%).

Notes

[16] Research space here is defined as the space used for sponsored R&D activities at academic institutions and for which there is separate budgeting and accounteing. Research space is measured in net assignable square feet (NASF). This is the sum of all areas on all floors of a building assigned to, or available to be assigned to, an occupant for a specific use, such as research or instruction. NASF is measured from the inside faces of walls. Multipurpose space that is partially used for research is prorated to reflect the proportion of time and use devoted to research.

[17] The S&E fields used in the National Science Foundation Survey of Science and Engineering Research Facilities are based on the National Center for Education Statistics Classification of Instructional Programs (CIP)—which is updated every 10 years (the current version is dated 2010). The S&E fields used in the FY 2011 Survey of Science and Engineering Research Facilities reflect the 2010 CIP update. Both the FY 2007 and FY 2009 surveys reflect the 2000 CIP standard. For a comparison of the subfields in the FY 2005 and FY 2007 surveys, see the detailed statistical tables for S&E Research Facilities: FY 2007. No major impacts on these data resulted from the CIP 2010 update.

[18] The science and technology field and subfield definitions were updated to the 2000 Classification of Instructional Programs starting with the FY 2007 Survey of Science and Engineering Research Facilities. Some of the observed declines in research space for health and clinical sciences and for physical sciences between FY 2005 and FY 2007 could reflect definition changes.

[19] Institutional sources includes an institution’s operating funds, endowments, private donations, tax-exempt bonds and other debt financing, and indirect costs recovered from federal and nonfederal sources.

[20] Only projects whose prorated cost was estimated to be $250,000 or more for at least one field of S&E were included.

[21] Because of rising capitalization thresholds, the dollar threshold for inclusion in the equipment category has changed over time. Generally, university equipment that costs less than $5,000 would be classified under the cost category of “supplies.”

[22] The “bricks and mortar” section of the Survey of Science and Engineering Research Facilities asks institutions to report their research space only. Therefore, the reported figures do not include space used for other purposes, such as instruction or administration. In the “Computing and Networking Capacity” section of the survey, respondents are asked to identify all of their cyberinfrastructure resources, regardless of whether these resources were used for research or other functions.

[23] Research-performing academic institutions are defined as colleges and universities that grant degrees in S&E and expend at least $1 million in R&D funds. Each institution’s R&D expenditures are determined through the National Science Foundation Higher Education Research and Development Survey.

[24] Academic institutions provided data on all computing systems with peak theoretical performance of 1 teraflop or faster. This defined the threshold for high-performance computing in the “Computing and Networking Capacity” section of the Survey of Science and Engineering Research Facilities. A teraflop is a measure of computing speed equal to 1 trillion floating point operations per second (FLOPS). FLOPS reflect the number of multiplications that a computer processor can perform within 1 second.

[25] These points have been cited as rationales for centralizing cyberinfrastructure and high-performance computing at several institutions (University of Arizona 2013; UCSD 2009; Bose et al. 2010).

[26] Clusters use multiple commodity systems, each running its own operating system with a high-performance interconnect network to perform as a single system. Massively parallel processors use multiple processors within a single system with a specialized high-performance interconnect network. Each processor uses its own memory and operating system. Symmetric multiprocessors use multiple processors sharing the same memory and operating system to work simultaneously on individual pieces of a program.

[27] Usable storage is the amount of space for data storage that is available for use after the space overhead required by file systems and applicable redundant array of independent disks configurations is removed. Online storage includes all storage providing immediate access for files and data from high-performance computing systems of at least 1 teraflop. Storage can be either locally available or made available via a network.