Optimal Overbooking Policy for Cloud Service Providers: Profit and Service Quality

A cloud federation is a current paradigm that enables partnered cloud providers to share idle capacities during low demand periods and to purchase spare resources during demand spikes. In this research, we propose an optimal overbooking policy to maximize federation members' profits and enhance cloud users' experiences. The proposed policy overcomes cloud providers' low utilization and increases their profits. Under the market-oriented cloud federation system, we use the number of idle resources in the cloud federation and the operational costs of those resources to help the provider decide on its instance exchange price. Under such a price mechanism, we develop an optimal overbooking model and identify the conditions necessary for optimal solutions. Through implementing the optimal mechanism, we observe that the proposed overbooking policy can improve a federated provider's profits and decrease the probability of service level agreement (SLA) violation. When a provider's capacity is relatively large and the provider adopts the proposed overbooking policy, it could achieve maximum profits and decline its SLA violation when it has unmet customer demands and there are idle resources in the cloud federation. Through establishing the cooperative game model of the cloud federation, we make a reasonable profit distribution based on Shapley value. The cloud provider's profits and the probability of the SLA violation change as the instance price, the distribution of unserved customers, the number of federation members and penalty cost change. Compared with the other overbooking policies and no overbooking mechanisms, our research improves profits and reduces cloud provider's overbooking risks, thereby presenting a win-win situation for both the individual providers and the cloud federation.

Abstract licence: CC BY-NC-ND

36 more references, whose wording the paper’s licence keeps private.

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.