Abstract:
Grid computing refers to a collection of compute resources from different geographic locations to achieve a common objective. It involves tremendous amounts of computatio...Show MoreMetadata
Abstract:
Grid computing refers to a collection of compute resources from different geographic locations to achieve a common objective. It involves tremendous amounts of computational task which require reliable resource sharing across the grid for efficiently executing the tasks. Load disparity is a major issue in grid computing. So effectively and efficiently utilize the potentials of huge distributed resources, load balancing algorithms are of primary importance. In this paper, we proposed an elegant approach using Bak-Tang-Wiesenfeld sandpile model for dynamically load-balancing of tasks across the processing sites in an energy efficient manner. The proposed approach reduces redundant migrations between processing sites by transferring tasks in a breadth first manner. In this approach avalanche propagation condition is stricter and depends on the average load of two adjacent layers of neighbours and use of signalling mechanism rather than actual transfer of tasks during an avalanche, until a destination comes in equilibrium. We experimentally validated by a number of numerical experiments using Google traces. The results of the experiments show the effectiveness of the proposed approach which reduces the number of migrations and improves the energy efficiency without much increasing the total flowtime and degrading the load balancing aspects as compared to state-of-the-art approaches.
Date of Conference: 18-20 December 2022
Date Added to IEEE Xplore: 28 March 2023
ISBN Information: