Nash equilibrium and social optimization in cloud service systems with diverse users

Cloud service systems typically rank the tasks from diverse users according to their privileges. Besides providing high performance service to membership-based users, extensive cloud service systems tend to offer service trials to normal users. In this paper, aiming to improve the utility and flexibility of cloud service systems, we propose an improved cloud architecture. In this architecture, membership-based users can accept the cloud service with high priority, while normal users accept the cloud service opportunistically. We build a multi-server queueing model with preemptive priority to analyze the stochastic behavior of the diverse users. On the basis of a two-dimensional Markov chain, we derive the steady-state distribution of the queueing model and give the mean sojourn time of normal tasks. By constructing revenue functions, we obtain the Nash equilibrium and the socially optimal arrival rates of normal tasks. Aiming to maximize the social revenue of the system, we present a pricing policy with an appropriate trial fee charged to normal tasks.