An online profit optimisation framework for diverse requests in distributed datacentres

Abstract

The rapid development of cloud computing attracts increasing amounts of enterprises and institutes to submit their diverse computing requests to the cloud service providers that are usually supported by large‐scale distributed datacentres. The explosion of service requests has drastically increased the operational expenditure of cloud service providers, which makes it imperative to find out a highly efficient method to reduce the cost and maximise the profit. However, the time and spatial differences of costs and the diverse characteristics of requests bring in multiple challenges to tackle this problem. In this work, we jointly take the electricity cost, communication cost and diverse characteristics of requests into consideration and propose a profit optimisation framework that includes three important decisions: service requests acceptance control, requests dispatching and resource provision. An efficient online algorithm is then designed to provide cloud service providers with the advice concerning these decisions to achieve the maximal time‐averaged profit over the long run. The rigorous theoretical analysis demonstrates that the proposed framework is able to approach a time averaged profit that is arbitrarily close to optimum. Extensive trace‐driven experiments further confirm the effectiveness and superiority of our framework. Copyright © 2016 John Wiley & Sons, Ltd.