Admission control and scheduling of remote processes in loosely-coupled distributed systems

Abstract

Scheduling and execution of remote processes are integral parts of designing grid and cloud computing systems, which aim to utilize the idle CPU and memory resources of a node. However, such execution tends to degrade the overall performance of a node through over-loading and delaying the execution of local processes. This paper proposes the Admission Control and Scheduling (ACS) algorithm for execution of remote processes in a node based on online estimation method. The experimental results illustrate that the algorithm successfully schedules a large set of CPU-bound and IO-bound remote processes without degrading overall performance of a node. The comparisons of CPU-utilization and memory-utilization of an idle node and a stable node executing remote processes show that, the average utilizations are enhanced by 26.65% and 24.5%, respectively. However, the ACS algorithm successfully maintains the load-balance in a node so that, the performance of the local processes remains unaffected. (C) 2013 Elsevier Ltd. All rights reserved.