As the HPCC Program reaches the middle of its initial five-year phase, the ARPA program is shifting focus from stimulating the development of the new scalable computing technology base and early experimental use toward developing the technologies needed to enable a broad base of applications and users, including their extension to a National Information Infrastructure. In addition, the foundations for future generations of computing systems involving even more advanced technologies are being developed.
The current scalable computing technology base is characterized by the first 100 gigaop class computing systems that are being experimentally used on a wide variety of problems in the scientific and engineering research communities. Scalable operating systems are enabling software development on high performance workstations connected to higher performance servers through networks.
The experience gained in the early experimental use of these new computing technologies is used to refine the next generation and guide the development of more advanced software and system development technologies. The combination of scalable computing and scalable networking technologies provides the foundation for solving both Grand Challenges with large scale parallel systems and National Challenges with large scale distributed systems. This enables HPCC to progress toward an NII.
ARPA is the lead DOD agency for advanced technology research and has the leadership responsibility for High Performance Computing (HPC) within DOD. The ARPA HPC Program develops dual use technologies with broad applicability to enable the defense and intelligence communities to build on commercial technologies with rapid development of more specific characteristics when needed. ARPA has no laboratories or centers of its own and executes its programs in close cooperation with the Office of Naval Research, the Air Force Office of Scientific Research, the Army Research Office, Service Laboratories, the National Security Agency, and other DOD organizations and Federal agencies. ARPA participates in joint projects with other agencies in the Federal HPCC Program, a variety of Defense agencies, the Intelligence community, and other Federal institutions.
Joint projects with other agencies are established to accelerate technology development and transition. ARPA joint projects with NSF include foundations for scalable systems, visualization, Grand Challenges, gigabit networks, and accelerating the maturation of systems software at NSF Supercomputer centers. Joint projects with NASA include an Internet software exchange, system software maturation, and ground stations for the ACTS gigabit satellite system. ARPA, NSF, and NASA also have a joint program in digital libraries. Joint projects with DOE include scalable software libraries and networking applications. ARPA is working with NIST to develop performance measurement technologies and techniques, privacy and trusted systems technologies, and the computer emergency response team system for the Internet. A joint project with NSA is developing gigabit network security technology and other secure and trusted systems technologies. In addition, a variety of early evaluation and experimental use projects involve different kinds of scalable parallel computing systems.
The ARPA program focuses on the advanced technology aspects of all five components of the HPCC Program as follows:
ARPA projects stimulate the development of scalable computing technologies that are capable of being configured as networks of workstations and large scale parallel computing systems capable of sustaining trillions of operations per second. Systems can be configured over a wide performance range. The systems will be balanced to provide the processor-to-memory, scalable interconnection, and input/output bandwidth needed to sustain high internal and external system performance. The modular design of the system units of replication will enable them to cover the full range from workstations to the largest scale distributed and parallel systems. Scalable systems with vector accelerators may be configured as parallel vector systems. Other kinds of accelerators such as field programmable logic arrays may be added for specialized applications. The largest scale parallel systems with hundreds to thousands of processors or more, are sometimes referred to as massively parallel systems. The input/output interfaces of these systems may be used to configure heterogeneous systems with high performance networks.
Scalable microkernel operating systems with a full complement of servers will enable software and system developers to work with a uniform set of application interfaces over the scalable computing base. Through the use of multiple servers, different application interfaces can be supported to enable the transition from legacy systems such as those available today. The system software may be configured as needed for particular applications including trusted and real-time systems.
Advanced components and packaging technologies including the associated design, prototyping, and support tools will enable higher performance and more compact systems to be developed. These technologies also enable the development of embedded systems so that computing can be put in specialized physical and environmental settings (such as airplanes, spacecraft, land vehicles, or ships).
Early evaluation and experimental use of new computing systems is an integral part of the overall development process. Policies and mechanisms have been developed that enable the timely purchase of new small to medium scale computing systems for the purpose of early evaluation and experimental use. As these technologies mature, larger scale systems are purchased and deployed by other parts of the HPCC Program in consultation with their user communities.
A multi-year Caltech project developed high performance interconnect and fine-grained parallel systems including boards consisting of 64 single nodes, scalable to thousands of nodes. Several commercial systems have adopted architectures based on this research, and are beginning to demonstrate how low cost modules can be configured to meet a broad range of applications.
ARPA projects develop scalable networking technologies to support the full range of applications from local networks, to regional networks, to national and global scale networks including their wireless and mobile extensions. Different kinds of communication channels, or "bitways", will be integrated to enable network connectivity to be achieved between users and their applications.
Internet technologies will be developed to enable continued scaling of the networks to every individual and system needing access. Scalable high performance networking technologies will be developed to enable gigabit speeds to be delivered to the end users. A variety of networking testbeds developed in cooperation with other agencies are used to develop, demonstrate, and experimentally deploy new networking technologies. As these technologies mature, larger scale systems are deployed by other parts of the Program in consultation with their user communities.
Experimental gigabit networks are overlaying and enhancing the Internet. ARPA's Networking Systems program develops and evaluates these technologies as foundations for a global scale, ubiquitous information infrastructure supporting Grand Challenge, National Challenge, and Defense needs.
ARPA projects develop the advanced algorithms and software technologies needed for computationally intensive applications. Scalable software libraries will enable optimal performance for common problem domains. Extensions to existing programming languages and the development of new ones will make it easier to program scalable systems. Optimizing compilers for individual processors and their scalable configurations including support for memory and process allocation will enable maximal performance for a given algorithm to be achieved. Design, analysis, visualization, and debugging tools will enable the development and support of scalable parallel and distributed computing systems through the use of common integrated frameworks. This will provide the foundation for the software and system development and support environments needed for computationally intensive problems such as the Grand Challenges. The application of HPCC technologies to specific Defense problems is performed by other ARPA or Defense programs.
Message passing, transparent access to user services, advanced memory management, and real time response are key to a new generation of portable microkernel-based operating systems. Coupled with scalable computing systems, these operating systems enable the introduction of multilevel caches, scalable device interfaces, and scalability to thousands of processors.
ARPA projects will develop the information infrastructure technologies needed for information and functionally challenging applications that emphasize information processing. Advanced Internet-based services will be developed to enable the effective deployment of distributed Internet-based systems including multiple media with access to high performance computing services and individual access points. Mobile and wireless technologies will enable users and their networks to access the information infrastructure with the appropriate authentication, privacy, and security.
Distributed data and object bases with transparent replication will enable the development and deployment of data, information, and knowledge repositories and services. A variety of access tools and interfaces will be developed to enable interactive access to the infrastructure. These will include intelligent functions such as knowledge based searching and alerting, planning and learning systems, and natural language understanding systems for speech, text, and images. This will provide the foundation for solving the large scale distributed, information and functionally intensive problems such as the National Challenges.
The three interconnected layers of the NII are a variety of "Bitways" for communication, services for information-based and computation- based resources, and National Challenge applications. Each layer focuses on protocols for common delivery mechanisms in order to insulate users from the details of the underlying technologies. In so doing, each layer sustains a diverse base of technologies and supports a broad base of suppliers. Communications protocols, for example, ensure delivery in a uniform manner; protocols for multimedia multicast and other services provide similar uniformity to applications developers; and applications developers use scalable open protocols to make rapid advances without major services reengineering.
The BRHR component is generally structured as an integral part of the other components. The ARPA program, along with NSF, provides the majority of Federal support to universities in computer and computational science. Fundamental limitations of HPCC technologies are identified and alternatives developed that will provide the basis for more advanced technologies in the future. An Historically Black Colleges and Universities program identifies individuals and groups that have the potential to contribute to the program. Small Business Innovation Research topics are formulated to provide another opportunity for small businesses to participate in the program. University projects provide an extraordinary opportunity to attract new people into the program, benefit from their talent, and prepare them for the new technologies. Basic research programs are often established in cooperation with NSF.
Develop system architectures for affordable computing systems, scalable from workstations to teraops systems.
Demonstrate scalable libraries, components, and tools to support software development on a scalable parallel system.
Develop and deploy gigabit networking protocols based on experimental testbed results.
Develop information infrastructure system architectures and early prototype services.