Data-Driven Stochastic Scheduling and Dynamic Auction in IaaS

Abstract

With the emergence of large scale data processing systems and big data analysis, cloud computing has become more and more popular. In this paper, we focus on the mechanism design in the infrastructure as a service (IaaS) cloud computing service market. Most of existing works on mechanism design assume static and independent individual utility, while in practice the cloud service is provided in a dynamic environment. To solve such problems, we propose a stochastic matching algorithm based on Markov Decision Process (MDP), which aims at optimizing the long-term system efficiency by considering the opportunity cost in the future. Based on the MDP formulation, we further design an efficient (EF), incentive compatible (IC), individual rational (IR) auction mechanism. Finally, we conduct experiment using Google cluster-usage traces dataset and show that the proposed MDP-based VCG auction mechanism can achieve EF, IC and IR properties simultaneously.