Latency-Aware Task Scheduling and Resource Selection using two phase Genetic Algorithm in Federation of Clouds

Abstract

Federation of cloud is a collaborative model in which multiple cloud providers participate by sharing resource, services and data across their platform. This collaboration is aimed to create a unified system where user can benefit from multiple cloud providers while maintain the independence of each participating cloud. One of the key component for sustainability and successful operation of a federation of clouds is its efficient resource selection model. Resource selection has always been a challenge in cloud computing and is even more challenging in a Federation or multi-cloud setup. Researchers have dealt with this problem in various ways. Most of these existing algorithms in consider processor and memory needs without considering the bandwidth requirements of an application.  In this paper, two-stage Latency-Aware Resource Selection (LA-RS) algorithm has been proposed to obtain a balance between various confronting objectives including Quality of Service (QoS), cost and completion time of applications. The first phase of the proposed algorithm figures out the top corresponding computing resources for the input tasks that satisfy their QoS requirements including cost and also considers network-latency in a federation or multi-cloud environment; the subsequent phase applies genetic algorithm that iteratively re-allocates the input tasks to optimize tasks execution time and cost. The comparison of proposed algorithm with existing algorithm clearly exhibits that along with considering the bandwidth of the underlying network, proposed algorithm achieves the objectives of optimal minimum execution time as well as optimal minimum cost.