Dynamic instance provisioning strategy in an Iaas cloud

Iaas cloud platforms can offer computational resources in forms of virtual machines according to the time-varying demands. The cost typically depends on the requirement of computation, storage and memory. Different resources provisioning scheme may result in significantly different cost. Therefore, it is a challenge to propose a dynamic instance purchasing scheme for minimizing the cost while guaranteeing quality-of-service (QoS). In this paper, a system model is designed for instance provisioning in Iaas cloud environment, and then a dynamic instance provisioning strategy based on the Central Limit Theorem is proposed to decide the minimum number of active instances for the upcoming tasks and guarantee the QoS demand which is defined as the overload probability. We evaluate our proposed methodology by using different simulations based on two real workload traces, and demonstrate that the proposed strategy can offer adaptive capacity provisioning and achieve dynamically resource allocation in an IaaS cloud.