November 23, 2004

November 11, 2005

Second Friday in November

annually thereafter

Computer Systems Research (CSR)

Computer systems are ubiquitous, and society is increasingly dependent on them. They range from microprocessors embedded in automobiles and appliances to worldwide grids of advanced processors, storage, graphics devices, and instruments interconnected by high-speed networks.  They are controlled by systems software, which has two main roles:  manage the underlying hardware resources, and provide abstractions and services that facilitate the implementation and execution of application programs.    However, too often computer systems fail, become compromised, or perform poorly. Moreover, they have become increasingly large and complex, thereby compounding problems.  Addressing these challenges requires major advances in systems software.

The Computer Systems Research (CSR) program supports innovative research and education projects that have the potential to: lead to significant improvements in existing computer systems by increasing our fundamental understanding of such systems; produce systems software that is qualitatively and quantitatively more reliable and more efficient; and/or, to produce innovative curricula or educational materials that better prepare the next generation of computing professionals.  The CSR program is also interested in projects that expand the capabilities of existing systems by exploiting the potential of new technologies or by developing innovative new ways to use existing technologies.  Projects supported will strive to make significant progress on challenging, high-impact problems—as opposed to incremental progress on familiar problems—and will have a credible plan for demonstrating the utility and potential impact of the proposed work.

The CSR program contains four topical areas:  embedded and hybrid systems, parallel and distributed operating systems, advanced execution systems, and systems modeling and analysis.   Projects may range in size from single investigators to teams of several investigators. 

The CSR program also accepts proposals for workshops and Small Grants for Exploratory Research (SGERs).

Computing platforms and applications are increasingly complex and dynamic.  They also cover a wide range of technologies and capabilities, such as globally distributed, heterogeneous platforms; high-end computing systems, including grids; local-area networks of enterprise systems; clusters of workstations; shared-memory multiprocessors; mobile and special-purpose embedded processors; real-time process controllers; and assemblies of embedded sensor systems and physical control actuation devices.

The Computer Systems Research (CSR) program supports innovative research and education projects that will lead to better computer systems by increasing our fundamental understanding of such systems and by producing better systems software.  The program aims to support projects that have the potential to lead to systems that are more reliable and robust, contain fewer bugs, have better and more predictable performance, provide useful new services, and exploit the potential of emerging technologies.

CSR also supports the development of innovative curricular materials that have the potential to greatly improve higher education on computer systems topics.  Such activities may be submitted as stand-alone projects or may be included as part of broader research and education projects.  Stand-alone curriculum development projects are expected to include strong justification of the need for the new materials and must include both plans for disseminating them to the community and for evaluating their effectiveness.

Systems software contains services and runtime support for applications as well as operating systems services for resource management and process control.  It interacts with applications from above and hardware and communication services from below.  In order to be effective, systems software must be developed with a systems view, where both its design and operation incorporate the effects of the entire computing system.  Projects that propose to create systems software thus have to consider the environment in which it will execute.  Projects that propose to model or analyze computer systems have to treat components not as isolated entities but with respect to their functionality and performance in relation to other components and layers of an entire system.  Creating new systems software will often require teams of researchers with expertise from several computing sub-disciplines.

The CSR program is organized into four interacting topical areas:

• Embedded and hybrid systems (EHS)
• Parallel and distributed operating systems (PDOS)
• Advanced execution systems (AES)
• Systems modeling and analysis (SMA)

The first three areas are organized around technologies, from embedded and real-time control systems, through multiprocessors and distributed systems, to high-end grids.  The systems modeling and analysis area addresses topics that cut across the other three areas.  In addition, several themes are common to the first three areas, such as resource and process management; interactions between systems software, applications, compilers, and runtime systems; and system services to support the composition and execution of applications.

The CSR program will fund proposals that address the topics in the four topical areas described below as well as projects that span areas.  The program is especially interested in projects that have the potential to lead to significant advances in systems software and that integrate research and education in creative ways.  Proposals are expected to describe how the project will demonstrate potential impact—e.g., by validating theories on real systems, by demonstrating the utility of new systems software and making it available to the community, and/or by disseminating and evaluating new educational materials or curricula.

Embedded Systems (EHS)

Information technology for embedded systems is a key accelerator of progress and innovation in modern engineered systems.  Embedded systems control devices that range from hearing aids and automobiles to the electrical power grid and global aviation infrastructure.  The nation’s critical infrastructures depend on embedded sensing and control systems.  However, current systems are built on top of decades-old technology, often as closed, single systems.  Future supervisory and real-time process control systems must be open and interoperable with other systems in a widely distributed environment. 

The EHS area supports research and education in scientific foundations and systems technology that will revolutionize the design and development of embedded and real-time systems.  EHS emphasizes temporal and hybrid (discrete and continuous) aspects of computational control for this class of systems.  The goals are to create and unify the foundations for interacting physical and computational systems and to supply technologies needed for building reliable embedded systems.  This requires a rethinking of the underpinnings, architectures, and system implementation philosophies for these kinds of systems.   A pervasive theme for the EHS area is the creation of high-confidence systems that integrate functionality with mechanisms that provide real-time, survivability, reliability, and security guarantees.  Key drivers for this theme are the coordination demands that are required by future generations of complex, networked embedded systems.  The area draws upon control theory, modeling, software generation, real-time and resource-aware scheduling, and formal methods.

Specific topics of interest to the EHS area include the following:

• Embedded systems software composition: algorithms, middleware, virtual machines, and system services; new concepts for distributed real-time and light-weight operating systems; component technology for functional and non-functional aspects of systems; and, programming methods for integration of embedded system services.
• Foundations and technology for distributed software control: hybrid discrete and continuous models for predicting behavior of systems; new concepts to support and secure future generation supervisory control and data acquisition (SCADA) systems and process control systems (PCS); and, scalable support for embedded sensors and sensor nets.
• Methods for modeling and design of embedded software and systems: foundations, design, implementation, synthesis, analysis, and certification methods; and, innovative approaches for failure modes, self-test, and recovery and reconstitution.
• Resource management and optimization: methods and tools for allocating, scheduling, and managing real-time, power-aware, distributed embedded systems; and, static and real-time dynamic scheduling technology that addresses and integrates multiple concerns such as real-time guarantees, power, clock frequency, thermal gain, RF emission and interference, network bandwidth, and criticality level.

Parallel and Distributed Operating Systems (PDOS)

Most computer systems are general-purpose meaning that they use commodity hardware and the systems software supports a variety of applications. This contrasts with embedded and real-time systems, which are closely tied to a single application and often use customized hardware.  General-purpose computer systems include uniprocessors, shared-memory multiprocessors, mobile devices and applications, local area distributed systems, clusters, wide area distributed systems, and computational grids.

The PDOS topical area supports research and education projects that advance the state of the art in operating systems software for the range of computer systems described above. The goals are to improve the capabilities, reliability, and efficiency of existing systems, to create new ways to utilize current technologies, and to harness the potential of emerging technologies.  Projects are expected to increase our fundamental understanding of how to design and build better operating systems and/or to create new types of systems and systems services and to demonstrate their utility through empirical prototypes that are evaluated and disseminated to the community.

Specific topics of interest to the PDOS area include the following:

• Resource management:  Scheduling, virtual memory management and protection; management of multiple levels of memory hierarchy and file systems; process and data migration; scalable and robust methods for communication and synchronization; virtualization of resources and efficient resource utilization; and, resource management across heterogeneous platforms with distinct administrative boundaries.
• System services:  Mechanisms that enable dynamic coalitions, such as peer-to-peer or ad-hoc groups; membership, naming, and authorization services; local and remote resource discovery and resource requests services; system monitoring for performance tuning or to provide resilience to faults; support for debugging large, widely dispersed distributed systems; checkpoint and recovery services; configuration management; and, customizable and adaptable systems services.
• System architecture:  New ways to organize systems, such as peer-to-peer; software architectures that scale to handle thousands of components; software architectures addressing changing trends, such as in sizes or speeds of processors, memory, address spaces, and backing store; multi-threaded kernel design and kernel-level management functions for end-to-end provisioning of services; and, dynamic, customizable kernels.
• System properties:  Fault-tolerance and reliability; efficiency; security; scalability; and, ability to cope with unexpected events.

Advanced Execution Systems (AES)

Advanced applications execute on large, heterogeneous high-end computing and grid platforms; require multiple resources, from fast processors to large data stores; and consume large amounts of communication network bandwidth.  These applications have dynamically changing resource requirements, and the underlying resources available to an application also change dynamically due to competing demands from other applications.  Consequently, an advanced execution environment has to support adaptive mapping of applications to underlying resources during execution. 

The AES topical area supports research and education projects that create systems software to facilitate the development and runtime support of complex applications.  One way to meet the above kinds of execution requirements is by means of runtime compiling systems and application composition systems.  These systems interface with the operating systems services developed in the parallel and distributed systems (PDOS) topical area, and they incorporate methods and tools developed in the systems modeling and analysis (SMA) topical area.

A runtime compiling system (RCS) extends the standard static notion of a compiler by embedding a portion of the compiler in the runtime system and endowing the RCS system with resource awareness and adaptive mapping capabilities.  In addition to analyzing an application’s characteristics, an RCS interfaces with the underlying operating systems services and resource managers in order to discover and request system resources and system services during the execution of the application.  An RCS also measures and queries features of components of the execution environment in order dynamically to optimize the mapping of the application to the underlying resources as resource availability changes or as the application’s needs change.  Developing RCS technology requires advances in areas such as programming models and tools for partitioning an application across distributed, heterogeneous computing platforms; new compiler techniques for determining functional and data dependencies across platforms and across multiple levels of memory hierarchy; application-level checkpointing and recovery; and mechanisms for matching an application’s resource needs to underlying resources when both are changing as the application executes.

An application composition system (ACS) allows an application to be constructed to fit the available resources and to adapt to changes in the underlying execution environment. It also supports application-level monitoring, debugging, and recovery.   An ACS interacts with and supports an RCS to compose applications dynamically.  Creating an ACS requires designing methods for automatically selecting application components; creating knowledge bases for application components; interfacing with the underlying computing platform models to determine suitable application components; and developing appropriate application component libraries and interfaces so the run-time portion of the RCS can link to such libraries.

System Modeling and Analysis (SMA)

The SMA topical area supports the development of methods and tools for modeling, measuring, analyzing, evaluating, and predicting the performance and correctness of complex computing and communications systems.  Ideally these methods and tools will be general and powerful enough to be applicable to the range of platforms addressed in this solicitation, from embedded systems to high-end computational grids.

Proposed methods and tools should support analysis of all levels of a system—hardware, systems software, and application—and of entire systems.  One way to do this is to create performance frameworks that employ models and measurements (in a plug-and-play fashion) of different levels of detail of each component and layer as needed for the specific analysis at hand.  Of specific interest are methods that provide hierarchical or multilevel analysis of such systems, enable assessment of the effects of individual hardware and software layers and components of these systems, and are pluggable into the performance frameworks to assess their impact on the performance of the entire system.  It is also important to be able to model how the behavior of the system scales as aspects of the application, systems software, or hardware change—e.g., when one shifts from a small prototype to a production environment or from one machine or to another with different architectural features.

Relation to Other Programs

Systems software exists in the context of applications, and it interacts with other software systems.  Consequently, many projects will address topics described above as well as topics covered in related programs.  If the major emphasis of a proposal is one of the types of systems or topics described in this solicitation, then it should be submitted to the CSR program.  If the major emphasis of the proposal is computer networks, then it should be submitted to the Research in Networking Technology and Systems (NeTS) program.  If the major emphasis is making systems more secure, then the proposal should be submitted to the Cyber Trust program.  A proposal that develops or uses compiler technology to address a system software problem should be submitted to CSR, but a proposal that addresses traditional compiler technology should be submitted to the Computing Processes and Artifacts (CPA) program.  Similarly, a project that develops programming languages or software tools to facilitate programming parallel applications should be submitted to CPA, whereas a project that develops systems software to support parallel computing should be submitted to CSR.  Finally, a project that develops middleware to support a specific computational science application or that hardens prototype systems software should be submitted to the NSF Middleware Initiative (NMI), whereas a project that proposes basic research on middleware services would be more appropriate for CSR.

The categories of proposers identified in the Grant Proposal Guide are eligible to submit proposals under this program announcement/solicitation.

An individual may appear as a PI, co-PI or Senior Personnel on no more than two proposals per annual CSR competition.

Limit on number of proposals:  None specified.

The CSR program will make two types of awards:

• Small.  These awards include one or two PIs, have budgets of up to $800,000, and durations of two or three years.   Approximately 35-45 small awards will be given in each annual CSR competition, with an estimated average award size of approximately $450,000.
• Team.  These awards include three or more PIs, have budgets of up to $2,000,000, and durations of three or four years.  Approximately 6-12 team awards will be given in each annual CSR competition, with an estimated average award size of $1,000,000.

If a project requires computing equipment or other infrastructure in order to be successful, and the required infrastructure is not available to the PIs, then it should be requested.  List the items on the budget pages and give a thorough justification in the budget explanation section of the proposal.  If the infrastructure request is for more than $100,000, the PI should discuss the need with one of the program officers identified in this solicitation.

The CSR program will also accept proposals for workshops and Small Grants for Exploratory Research (SGER).  These may be submitted at any time, but a prospective PI is required to discuss their idea with one of the program officers mentioned in this solicitation before submitting such a proposal.

In unusual circumstances, the Division of Computer and Network Systems will entertain proposals that are beyond the scope and funding levels noted elsewhere in this solicitation.  Such proposals would be expected to explore groundbreaking or paradigm-changing ideas or to pursue a grand challenge requiring the work of a substantial number of researchers.  Projects of this type might well include multidisciplinary investigators and cover topics in more than one CNS or CISE program.  PIs who have in mind such a project must first brief the appropriate program officers and the CNS Division Director.  PIs may submit a full proposal only after being given permission to do so.  The briefing must take place before the program solicitation deadline so the Division can plan for the receipt and review of this type of proposal. 

Estimated program budget, number of awards, and average award size/duration are subject to the availability of funds.

The following information deviates from the GPG guidelines:

To assist NSF staff in sorting proposals for review, proposal titles should begin with “CSR---key” where key is an acronym for one of the topical areas.  If a proposal addresses topics in more than one area, it should be directed at whichever one of the areas is most appropriate.  Use the following acronyms to identify the topical area:

• EHS --- embedded and hybrid systems
• PDOS --- parallel and distributed systems
• AES --- advanced execution systems
• SMA --- system modeling and analysis

For example, a proposal might have a title such as “CSR---PDOS: Resource Management in Peer to Peer Systems.” 

A workshop proposal should have the key “WORKSHOP” and a SGER proposal should have the key “SGER.”

Every proposal must include a discussion of broader impacts.  Appropriate goals for the broader impacts component include the integration of education and research, promoting diversity in the computer systems workforce, developing substantial experimental research educational experiences, and developing curriculum and supporting materials in emerging computer systems areas.  The following URL contains several examples of broader impacts activities: https://www.nsf.gov/pubs/2002/nsf022/bicexamples.pdf.

November 23, 2004

November 11, 2005

Second Friday in November
annually thereafter

Systems software exists in the context of applications, and it interacts with other software systems.  Consequently, many projects will address topics described above as well as topics covered in related programs.  If the major emphasis of a proposal is one of the types of systems or topics described in this solicitation, then it should be submitted to the CSR program.  If the major emphasis of the proposal is computer networks, then it should be submitted to the Research in Networking Technology and Systems (NeTS) program.  If the major emphasis is making systems more secure, then the proposal should be submitted to the Cyber Trust program.  A proposal that develops or uses compiler technology to address a system software problem should be submitted to CSR, but a proposal that addresses traditional compiler technology should be submitted to the Computing Processes and Artifacts (CPA) program.  Similarly, a project that develops programming languages or software tools to facilitate programming parallel applications should be submitted to CPA, whereas a project that develops systems software to support parallel computing should be submitted to CSR.  Finally, a project that develops middleware to support a specific computational science application or that hardens prototype systems software should be submitted to the NSF Middleware Initiative (NMI), whereas a project that proposes basic research on middleware services would be more appropriate for CSR.