Adaptive Task Scheduling and Resource Management for Managing Flash Crowds in Cloud Environment

Abstract

Cloud computing being a master in controlling wide variety of virtual resources, incorporated scheduling which imprinted its footsteps deeply. For every job, multiple such virtual resources from cloud will be utilized. Manual scheduling of resources for each job results in complexity and as such remains an impractical solution. During Resource Allocation (RA), a node’s failure could cause interruption of cloud service. Present RA techniques struggle to achieve high throughput in less execution time. To handle flash crowds thereby serving the aforementioned challenge, an appropriate task scheduling algorithm in addition to RA techniques is essential. This work focuses on job scheduling and effective resource allocation for flash crowds that used online streams of multiple user requests as input. During RA, the exact split of data related to a specific user request was recognized and preprocessed. The Closed Frequent Itemset (CFI)from the keywords related to the corresponding query were obtained followed by computation of their corresponding entropy values. Then, the scheduling process is done using Normalized K- means Algorithm (NKMA) and firefly algorithm with respect to the obtained entropy values. Finally, Using Genetic algorithm based on Cauchy Mutations (CMGA) appropriate resources were allocated to the scheduled tasks. The experimental evaluation of the proposed work demonstrates that efficient cloud-based resource allocation and job scheduling can be accomplished, resulting maximum throughput and reduced execution time.