Algorithms for Job and Resource Discovery for the Meta-Scheduler of the DEISA Grid

This paper describes a distributed meta-scheduling architecture for the DEISA grid in which site-local proxies migrate jobs between the participating sites with the aim to improve user- and administration-defined goals like job runtimes, hardware utilization, load balancing, etc. The meta-scheduler employs P2P-based algorithms to discover available resources for its local jobs and jobs for its available resources respectively and thereby achieves continued operation even in the case of network or site failures.