Dynamic SLA based elastic cloud service management: A SaaS perspective

Elasticity is one of the most attractive features of cloud computing. While elastic resource requirements of a cloud service user often exhibit periodic patterns, the absence of a mechanism to specify precisely the elastic resource requirements forces the cloud service provider to either provision unbounded resources or offer services with rigid and static resources, none of which is an efficient or economic proposition to the tenant. We propose, in this paper, a novel model of dynamic service level agreement (SLA) for the tenants to specify their elastic resource requirements so that the cloud service provider can take an informed decision for resource management, thereby optimizing on resource utilization, user expenses and generated revenue. We study the problem from a Software-as-a-Service (SaaS) perspective. We propose an intuitive model of dynamic SLA that can capture the anticipated license requirement variation of an SaaS user. We also propose ILP and greedy-heuristic based optimization approaches for the service provider to solve the tenant on-boarding problem with dynamic SLA constraints. Experimental results show that our solution produces optimal or near-optimal (for heuristic based approach) performance in terms of overall resource utilization and the economy for both the service provider and the tenants.