Fault Tolerant Design in the Earth Observing System Archive

The Archive for the Earth Observing System (EOS) is one of the largest and highest data rate archives in the world. The EOS Archive is referred to as EOS Core System (ECS) and is a multi-site distributed data warehouse of Earth-oriented satellite images and science algorithms/reports. Its data holdings are projected to approach five petabytes by 2002. Each distributed site is referred to as a “Distributed Active Archive Center” or DAAC. The DAAC sites are being incrementally delivered with final deployment by the end of 2000. One of the sites, the EROS Data Center (EDC) in South Dakota, is receiving and archiving Landsat data in addition to the data generated by the instruments on the Terra satellite launched in December of 1999. Four of the DAACs will begin receiving Terra data in early 2000 [1]. The ECS archive architecture is based on a multi-site, distributed, client-server model. Its components are interdependent. As in any large and reasonably complex system robustness and ability of functional components to recover from faults is of great importance. In particular, ECS places heavy emphasis on data integrity and data capture robustness. This paper briefly describes the design of the hardware and software to insure the EOS data is captured and distributed in spite of faults. The description of hardware failover is confined to the Ingest Component design. The paper is intended as an introduction to the Poster Presentation material, and other components are discussed in the Poster Presentation itself.