Level Based Task Prioritization Scheduling for Small Workflows in Cloud Environment

Abstract

Cloud computing is aimed at delivering computing services as a utility. One of the Cloud computing services is Workflow as a Service (WFaaS). Workflow scheduling is a vital area in WFaaS. The objective of this paper is to propose a scheduling algorithm that meets the Quality of Service constraints such as Makespan and Cost. The proposed algorithm Level Based Task Prioritization (LBTP) follows the list scheduling strategy. The LBTP algorithm alters task prioritization phase and the resource selection phase, where three different algorithms like Earliest Finish Time, Parent Resource Allocation and Round Robin are applied. The LBTP algorithm is tested for small workflows. The performance analysis is done by varying the Communication to Computation Ratio and number of tasks. The scheduling metrics, Makespan and Cost of the proposed algorithm are compared with the Customer Facilitated Cost based Scheduling (CFCSC) algorithm. The experimental results show that Earliest Finish Time resource selection procedure outperforms CFCSC algorithm with respect to Makespan for communication intensive graphs. For computation intensive graphs, the parent resource allocation procedure produces better Makespan. With respect to Cost parameter, irrespective of total number of tasks in the Direct Acyclic Graph, all the resource selection procedures have incurred minimum cost. The proposed algorithm helps the Cloud Provider to select the scheduling algorithm in accordance with the Quality of Service parameters. Further, the LBTP algorithm can be tested with Regular Scientific workflows like Montage and Cybershake.