Resilient service provisioning in cloud based data centers

Abstract

Cloud service providers usually have several geo-distributed data centers to meet these increasing demands. On peak times, these data centers receive unpredictable amount of workload that might go beyond their capacity and service providers need to distribute this workload with the objective of maximizing their revenue. This objective is hard to be guaranteed given the possible incidents that may cause cloud services interruption such as power outages, security breaches, natural disasters, etc. Since cloud providers are facing serious threats to the continuation of their services offering which will cause a serious decrease in their revenues, they must have proper strategies for handling such incidents. While a rich volume of recent research works focused on optimizing cloud data centers operations in order to reduce their operational cost (OPEX), little attention has been paid to such serious incidents and how they would affect service provisioning. In this study, we use mixed integer-linear programming to model the problem of maximizing revenue of a given workload by creating/expanding/renting data centers. The model takes into account a wide range of incidents affecting the users traffic and/or the data centers service capacities. Moreover, it handles many issues such as different types of costs (capital, operational and recovery), power consumption, carbon footprint, SLA related issues such as constraints (e.g., delay) and penalties, etc. To the best of our knowledge, this formation is unique among the body of existing works. Finally, to evaluate the proposed model and the impact of different parameters on its performance, several simulation experiments are conducted.