Variability in Behavior of Application Service Workload in a Utility Cloud

Abstract

Using the elasticity feature of a utility cloud, users can acquire and release resources as required and pay for what they use. Applications with time-varying workloads can request for variable resources over time that makes cloud a convenient option for such applications. The elasticity in current IaaS cloud provides mainly two options to the users: horizontal and vertical scaling. In both ways of scaling the basic resource allocation unit is fixed-sized VM, it forces the cloud users to characterize their workload based on VM size, which might lead to under-utilization or over-allocation of resources. This turns out to be an inefficient model for both cloud users and providers. In this paper we discuss and calculate the variability in different kinds of application service workload. We also discuss different dynamic provisioning approaches proposed by researchers. We conclude with a brief introduction to the issues or limitations in existing solutions and our approach to resolve them in a way that is suitable and economic for both cloud user and provider.