DOI: 10.5176/978-981-08-7656-2 A-07
Authors: Andy S. Chiou, Wen-Jiun Lin
Abstract: The emerging Cloud and Grid computing have become the de facto distributed computation. These environments are dynamic in the sense that the arrivals of user requests are unpredictable and due to this versatility, traditional scheduling algorithms fail to match user requests to the most feasible processing elements because no prior knowledge about the requests is available until they arrive at the schedulers. The
situation is getting worse particularly in a large scaled system, in which the single scheduler is often the bottleneck that the queue is congested with a large amount of user requests. One of the solutions to this problem is to use multiple schedulers. However, the scheduling complexity can rise dramatically since a scheduler must consult others and maintain a global state of all processing elements in order to generate a feasible plan. Sometimes the
imposed costs of synchronizing the schedulers and processing elements can turn into the major cause of system degradation. In this research, we take the resource competition approach in which the schedulers compete for the resources without acknowledging other competitors and the processing elements accept only the first request arrives and reject the others. We conduct thorough simulations and the results are positive.
Keywords: Grid; Cloud Computing; Resource Competition;
Distributed Scheduling
Updating...