The design of a petabyte archive and distribution system for the NASA ECS project

The NASA EOS Data and Information System (EOSDIS) Core System (ECS) will contain one of the largest data management systems ever built - the ECS Science and Data Processing System (SDPS). SDPS is designed to support long term Global Change Research by acquiring, producing, and storing earth science data, and by providing efficient means for accessing and manipulating that data. The first two releases of SDPS, Release A and Release B, will be operational in 1997 and 1998, respectively. Release B will be deployed at eight Distributed Active Archiving Centers (DAAC's). Individual DAAC's will archive different collections of earth science data, and will vary in archive capacity. The storage and management of these data collections is the responsibility of the SDPS Data Server subsystem. It is anticipated that by the year 2001, the Data Server subsystem at the Goddard DAAC must support a near-line data storage capacity of one petabyte. The development of SDPS is a system integration effort in which COTS products will be used in favor of custom components in very possible way. Some software and hardware capabilities required to meet ECS data volume and storage management requirements beyond 1999 are not yet supported by available COTS products. The ECS project will not undertake major custom development efforts to provide these capabilities. Instead, SDPS and its Data Server subsystem are designed to support initial implementations with current products, and provide an evolutionary framework that facilitates the introduction of advanced COTS products as they become available. This paper provides a high-level description of the Data Server subsystem design from a COTS integration standpoint, and discussed some of the major issues driving the design. The paper focuses on features of the design that will make the system scalable and adaptable to changing technologies.