A LOAD BALANCING MODEL USING FIREFLY ALGORITHM IN CLOUD COMPUTING

Abstract

Cloud computing is a model that points at streamlin ing the on-demand provisioning of software, hardware and data as services and providing end-users with f lexible and scalable services accessible through th e Internet. The main objective of the proposed approa ch is to maximize the resource utilization and prov ide a good balanced load among all the resources in cloud servers. Initially, a load model of every resource will be derived based on several factors such as, memory usage, processing time and access rate. Based on t he newly derived load index, the current load will be computed for all the resources shared in virtual ma chine of cloud servers. Once the load index is computed f or all the resources, load balancing operation will be initiated to effectively use the resources dynamica lly with the process of assigning resources to the corresponding node to reduce the load value. So, as signing of resources to proper nodes is an optimal distribution problem so that many optimization algo rithms such as genetic algorithm and modified genetic algorithm are utilized for load balancing. These algorithms are not much effective in providin g the neighbour solutions since it does not overcome exploration and exploration problem. So, utilizing the effective optimization procedure instead of genetic algorithm can lead to better load balancing since it is a traditional and old algorithm. Accordingly, I hav e planned to utilize a recent optimization algorith m, called firefly algorithm to do the load balancing o peration in our proposed work. At first, the index table will be maintained by considering the availability of virtual servers and sequence of request. Then, l oad index will be computed based on the newly derived f ormulae. Based on load index, load balancing operation will be carried out using firefly algorit hm. The performance analysis produced expected results and thus proved the proposed approach is efficient in optimizing schedules by balancing the loads. The average time obtained for the proposed approach is 0.934 ms.