An energy-efficient incentive mechanism for resource recycling in multi-tenant datacenters

Abstract

Multi-tenant public cloud is an important sector of cloud computing. While it has witnessed an unprecedented proliferation in the past years, its low resource utilization that incurs enormous energy waste has received great attention. Since the resource of multi-tenant datacenters is typically rented to the tenants in a reserved manner, the cloud provider is unable to reduce the energy consumption or improve resource utilization via turning down idle servers. To address this challenge, this paper proposes an efficient incentive mechanism to stimulate the recycling of reserved idle resources. Specifically, by economically compensating the tenants, the reserved resources that are idle can be recycled by the cloud providers to host other cloud applications, such as e-commerce and search engine. Taking advantage of the Lyapunov optimization method, this paper rigorously designs and analyzes a dynamic pricing scheme for the recycled reserved resources. Without requiring the future information as a priori, the proposed pricing scheme can make online decisions to optimize the long-term operational cost of the cloud provider, while still maintaining the stability of the system. Both rigorous theoretical analysis and extensive trace-driven simulations verify the efficacy of the proposed incentive mechanism.