Distributed Metadata with the AMGA Metadata Catalog

Abstract

Catalog Services play a vital role on Data Grids by allowing users and applications to discover and locate the data needed. On large Data Grids, with hundreds of geographically distributed sites, centralized Catalog Services do not provide the required scalability, performance or fault-tolerance. In this article, we start by presenting and discussing the general requirements on Grid Catalogs of applications being developed by the EGEE user community. This provides the motivation for the second part of the article, where we present the replication and distribution mechanisms we have designed and implemented into the AMGA Metadata Catalog, which is part of the gLite software stack being developed for the EGEE project. Implementing these mechanisms in the catalog itself has the advantages of not requiring any special support from the relational database back-end, of being database independent, and of allowing tailoring the mechanisms to the specific requirements and characteristics of Metadata Catalogs.