Investigating a Benchmark Cloud Media Resource Allocation and Optimization - TI Journals

Abstract

Abstract: The essence of high performance computing systems emerged with the extension of big data issues and it entailed to cloud and other HPC systems to come into existence. A challenging issue in designing such systems is resource allocation, which explicitly affects the total efficiency of system by involving factors like the number of nodes, geographical distance, communication latencies, etc. In this study, authors will take a look at resource allocation issue in HPCs especially in clouds. Due to it, after a short introduction, different challenges are investigated. Then, one of the most common methods of resource allocation through resource broker is probes. Resource broker is an entity that allocates the resources to the applicants based on the policies of the system. For instance, it might allocate the resources with the lowest price to the applicants with the highest budgets. Prices and budgets are also calculated according to the policies of the system, e.g. based on the media resource’s computing ability, media resource’s lowest and highest prices, etc. For this purpose, resource broker constructs probability matrix P and do the resource allocation based on it. Also, the way of optimizing the matrix P is explained. Results of this method will be the balancing the workload of resource nodes with the number of applicants.