Economics models and policies for cloud federations

Abstract

Cloud federation has emerged as an effective solution offering worldwide coverage, dynamic infrastructure scaling and improved QoS for the demanding cloud services. In this paper, we present a model for Cloud Service Providers (CSPs) federation and we investigate the economic benefits of CSPs under different federation modes. Each CSP is modeled as an M/M/1 queue with arrivals corresponding to computation tasks and service rate that captures the computational capabilities of the CSP. Each CSP earns revenue by charging its customers according to a QoS-dependent pricing function, and it undergoes a cost due to energy consumption of its infrastructure. We propose a model for the formation of cloud federations, according to which each CSP may forward part of the workload stemming from its customers to other CSPs. We define three federation modes with varying degrees of CSPs' interaction, namely the strong, weak and elastic federations. In strong federations, the CSPs jointly decide on their forwarding policies to maximize the total profit of the federation; then, they share these profits according to certain profit sharing policies. In weak federations, a game arises, in which each CSP follows a forwarding policy that aims to maximize its individual payoff, which however incorporates some fairness. In elastic federations, each CSP again aims to maximize its individual payoff, but it has the freedom to tune the degree of its selfishness through a pricing function. The numerical results validate and quantify the conjecture that federation can incur substantial monetary benefits and achieve a near to optimal QoS.