[an error occurred while processing this directive]

NERSC 3 Greenbook

Next: Scientific Applications: Third Party Up: Software Previous: Libraries, Compilers, Debuggers, and

Distributed Computing and Unified Production Environments

With the advent of software infrastructure tools supported by the vendor community such as the Distributed Computing Environment (DCE), one new challenge to, and opportunity for, the ER user community is to harness the power of physically distributed computational resources. The development of tools for distributed computing by the research community is also well underway as evidenced by the I-WAY demonstrations performed at Supercomputing 95. The Distributed Computing Coordinating Committee (DCCC) is providing various infrastructure tools for the DOE-ER community. The challenge for the ERSUG community is to help focus DCCC efforts on tools that will enable ER scientific efforts and application development.

NERSC is planning a Unified Production environment to facilitate the interaction of the various DOE computational communities. The basic goals for the Unified Production environment are:

• The centralized NERSC environment must serve all the high-end needs. No local user environment, no matter how rich, can hope to offer peak computational speeds of on the order of 200-500 GFlops or storage capabilities of on the order of 200 TByte. In short, a centrally located capability engine (MPP, cluster of SMPs or PVPs, or heterogeneous system) within a Unified Production Environment, with adequate internal balance in support infrastructure (disk, tertiary storage and complementary high-end capability shared memory platforms) remains a major goal.
• The NERSC environment must continue to offer all the services required by the high-end user with limited local resources: this includes code development, computing and assimilation capabilities, as well as information services and consulting. Information services, consulting and collaborative ventures with NERSC staff are becoming more important because of the complexity in the new programming and assimilation environments. Without expertise, the capability latent in the hardware will not be realized.
• Ancillary services must be easily accessible, and must be offered on the appropriate platform. The MPP is not currently a viable platform for code development; NERSC must facilitate the user's migration through all the phases of computational research projects by providing a system administration environment that makes this easy, and that allows the user to select the platform most appropriate to the computational task at hand. That is what is meant by the Unified Production Environment.
• For users with a rich local environment, the Center must strive to act as the high-end complement to the local environment. Many of the aspects of unification practiced within the centralized NERSC environment can be practiced within the NERSC - user (that is, centralized - remote) environment. Work in this area could begin with one or two pilot projects involving remote sites.
• As scientific collaborations have increased in size the role of distributed computing and a unified or seamless environment has grown in importance. The centralized environment at NERSC needs to be both compatible with resources at other large (centers) and small (desktop) sites in terms of application software and security services. Additionally effective network connections for data access between these sites is crucial to the scientific missions.

NERSC 3 Greenbook

Next: Scientific Applications: Third Party Up: Software Previous: Libraries, Compilers, Debuggers, and