1 A Perspective on Metacomputing

The goal of a metacomputing system is to increase job throughput by using all available resources in a grid, while transparently providing for resource heterogeneity, network heterogeneity, and a security model for the multiple administrative domains that correspond to the autonomous sites comprising the underlying grid. The initial vision of metacomputing was to harness idle cycles of a local network, and grew to encompass those in geographically distinct government laboratories. Such systems were not without problems: each cluster had different hardware, resulting in non-portable and non-scalable software and preventing the realization of infinitely-scalable computing. In recent years several new systems have been developed with a focus on developing working systems that provide a comprehensive set of resources. We examine a representative subset of these systems in this paper, as well as some of the conceptual problems stemming from the inhomogeneities in a metacomputer