Computer Science
Computer
science includes the systematic study of computer hardware and software,
computer methodologies, and methods for testing system performance. There
is a good deal of interdisciplinary overlap between computer science
and applied
mathematics, especially in the area of numerical methods and parallel
algorithms.
CISL conducts research in computer science to provide the best computational facilities possible to the NCAR/UCAR scientific community. These activities include tracking new computer technologies, benchmarking for the geosciences, and using computer simulations to evolve and expand the capacities of NCAR's Mass Storage System.
Tracking new computer technologies
CISL tracks emerging computer technologies to ensure the efficient use of current computing resources and the selection of the most appropriate computers for NCAR/UCAR's future computational needs. CISL works with the scientific community to formulate and evaluate the system requirements that will meet the scientific objectives of the institution.
CISL assesses capabilities of new systems available from vendors and plays an active role in evaluating programming languages and paradigms. To keep up with the latest developments, CISL staff review computer literature, attend technical conferences and vendor presentations, run benchmarks, and research experimental systems.
Periodically CISL executes open, competitive procurements, soliciting vendor proposals and evaluating them on a best-value basis, taking into account technical, business, and price factors to maximize the benefit to the researchers who will utilize the system.
Benchmarking for the geosciences
Because divergent components and architectures make it difficult to compare the performance of high-end computer systems simply by checking technical specifications, CISL uses a series of standardized tests, called benchmarks, to assess performance characteristics. Benchmarks are designed to reproduce a specific type of workload on a component or system. CISL uses benchmarks to evaluate emerging parallel architectures, commodity clusters, vector supercomputers, system interconnects, and other technologies to determine their suitability for geoscience applications.
Simulating the Mass Storage System
NCAR's Mass Storage System (MSS) is an archive of more than 2 petabytes of computational analyses and observational data used for atmospheric research and related geosciences. Its sheer size and complexity makes it difficult to do capacity planning and performance tuning. However, CISL has developed a simulation program that models MSS components such as tape drives, data silos, disk subsystems, and software to predict the impact of system changes on MSS performance. This allows CISL to research and virtually experiment with different hardware and software configurations before actually making changes or purchasing new equipment, ensuring cost savings and optimal design and functioning for the MSS as it evolves.
