Resilient and performance-guaranteed composite services in cloud computing

Abstract

Services offered by different cloud service providers can be federated into composite services that can execute a variety of tasks subject to a service level agreement (SLA). An SLA is defined as quality of service (QoS) metrics and a fee. In this paper, we consider the QoS metrics: trustworthiness, percentile response time, and vulnerability. The percentile response time metric is a statistical bound stating that the end-to-end response time is less than a given bound for a certain percent of time. We first describe an efficient and accurate numerical solution for the calculation of the percentile response time. Then, we consider all these QoS metrics in a resilient and performance-guaranteed cloud service composition problem that typically arises in inter cloud computing. We present a framework for tackling the SLA-guaranteed cloud service composition problem in which an optimisation problem subject to QoS constraints is solved using an efficient numerical approach.