Cooperating runtime systems in LiPS

Abstract

Performing computation using networks of workstations is increasingly becoming an alternative to using a supercomputer. This approach is motivated by the vast quantities of unused idle-time available in workstation networks. Unlike comptuting o a tighty coupled parallel computer, where a fixed number of processor nodes is used within a computation, the number of usable nodes in a workstation network is constantly changing over time. Additionally, workstations are more frequently subject to outages, e.g. due to reboots. The question arises how applications, adapting smoothly to this environment, should be realized. LiPS is a system for distributed computing using idle-cycles in networks for workstations. This system is ints version 2.3 is currently used at the Universität des Saarlandes in Saarbrücken, Germany to perform computationally intensive applications in the field of cryptography on a net of approximately 250 workstations and should be enhanced to work within an environment of more than 1000 machines all over the world within the next years. In this paper we present the runtime systems of LiPS along with performance measurements taken with the current LiPS development version 2.4.