Performance Tuning of a High Capacity/High Performance Archive for the Earth Observing Systems Project

Abstract: The Archive for the Earth Observing Systems (EOS) Project reflects the scale of the project itself. Its data holdings over all the Distributed Active Archive Centers (DAACs) are projected to exceed two Petabytes by the year 2002. This paper describes the experience and results of the integration and tuning process for the hardware and Commercial Off The Shelf (COTS) software used today in the EOS Core System (ECS) archive. Sizable performance improvements have been realized to date through the combined efforts of the vendors and the tuning team. The process is still continuing and, indeed, will continue through the life of the archive. The effort to date centered around improving the throughput for the individual data streams to the archive tape drives and the cumulative throughput to all the drives. The work was specifically directed towards the COTS hardware and software integration and tuning adjustments. Throughput performance improvement via efficient data organization on tape was not addressed during this effort, but will be in the future work.