International Journal of Computer Applications |
Foundation of Computer Science (FCS), NY, USA |
Volume 126 - Number 14 |
Year of Publication: 2015 |
Authors: Divya Diwakar, Sushil Chaturvedi, S.K. Shrivastava |
10.5120/ijca2015906306 |
Divya Diwakar, Sushil Chaturvedi, S.K. Shrivastava . Water-Filling: A Novel Approach of Load Rebalancing for File Systems in Cloud. International Journal of Computer Applications. 126, 14 ( September 2015), 28-33. DOI=10.5120/ijca2015906306
File systems serves as the backend for cloud computing and load balancing is the relevant issue in context of resource utilization for distributed file systems in cloud. Prior to this, it is fruitful to identify the load on the storage servers (nodes) which is equivalent to number of file chunks it stored. Here is an extension of load balancing i.e. water-filling load rebalancing operated on distributed approach based on water-filling methodology, contrasting all the earlier algorithms that were grounded on centralized and distributed approaches, is used for balancing the load on servers by distributing file chunks making it more amplified to perform map reducing tasks. Water-filling approach enhances the scope of algorithm by calculating the total load exchange cost and rejoining cost in terms of file chunks migrated. Besides this, distributed approach, which employs self reliant load balancing on each node, is preferred due to its effortlessness. In distributed approach the node having highest and the lowest load is preferred to exchange chunks but often not on least possible load exchange cost. In this paper, an improved load distribution task based on physical network locality significance is calculated by water-filling algorithm, is used as metric for minimizing the load exchange cost to improve the load balancing for overcoming the shortcomings of centralized and distributed approach. Experimental results reports that water-filling load balancing algorithm is 81% better in load distribution than distributed load rebalancing, coagulates less load movement cost and even predicting reduced rejoining cost for migration of chunks in the panoptic environment of cloud.