A Makespan Based Framework for Detection of SLA Violations in Cloud Computing Environment

Abstract

Service level agreement (SLA) plays a vital role in delivery of cloud computing services. SLA is an agreement between the cloud service provider and the customer(s). It specifies the quality of services (QoS) to be delivered along with time deadlines and penalties to be imposed, in the event of a violation of agreement. There are research evidences that show that an agreement with larger number of check metrics may result better QoS. Servers downtime, minimum bandwidth, maximum computation time, etc., are some examples of these metrics. The work presented here introduces a new metric called makespan in SLA and proposes a framework for detection of violations in SLA. Makespan is the maximum completion time of all jobs submitted to the cloud. The proposed framework mathematically calculates the makespan at the time of SLA definition. The time period, for which SLA is violated, can be estimated from the comparison of simulation results. We also calculated makespan for different cloudlet scheduling algorithms. The study also performed comparative analysis of popular scheduling algorithms like first come first serve (FCFS), shortest job first (SJF), round-robin (RR), and particle swarm optimization (PSO). The experimental work was carried on CloudSim simulator, and algorithms were implemented in JAVA programming language. The results obtained are promising. Following are some of the key findings of the study which states that for small number of cloudlets, SLA violations in SJF are highest as compared with other algorithms. For large number of cloudlets, SLA violations in RR are highest. PSO has shown uniformly low SLA violations for both small number and large number of cloudlets. The study not only proposes a novel technique for detection of SLA violations but also highlights possible research extensions in this field.