Abstract
Conference Title: 2017 Computing Conference Conference Start Date: 2017, July 18 Conference End Date: 2017, July 20 Conference Location: London, United Kingdom In cloud environments, the use of data-intensive applications is becoming more common than ever. For efficient computation, data-intensive applications require a bigger volume of data. Approaches that are aimed at intensive computation, such as conventional meta-scheduling approaches, do not consider the data requirement of the application, which leads to poor performance. One of the big challenge in Cloud environment is scheduling the application data in an efficient manner. Additionally, average turnaround time and process utilization should be considered in such environments. An evaluation of a cloud scheduling strategys performance is not complete without taking into account the worst case turnaround time. In this paper, we proposed a VMs placement algorithm which takes into account three important factors that affect job completion time. These factors are data locality, processing time and queuing time. In the proposed scheme, the data transfer is taken as an important factor in the computation and is thus explicitly considered in the scheduling process. Additionally, the proposed algorithm is covered the wasting time during jobs queuing by applying overlap technique. In this technique the job data will transferring in advance during it queuing and other jobs in the processing stage. This could save some time and improve the application performance. Based on the simulation results, which shows that our proposed algorithm brings significant benefits to jobs that are data sensitive.