Privacy-preserving coordinated operation of cross-enterprise data centers

As critical digital infrastructures, data centers are often built with idle computing resources to reliably support large amounts of computational workload. The utilization of those idle resources is restricted by affiliations of data centers and can be enhanced if we allocate workload coordinately to data centers of different enterprises. Besides, conventional centralized optimization, collecting detailed information of data center status as boundaries, cannot be adopted here due to privacy concerns. Considering these issues, this paper presents a privacy-preserving coordinated operation tailored for cross-enterprise data centers. This approach integrates the alternating direction method of multipliers (ADMM) and homomorphic encryption within a distributed algorithm, ensuring privacy-preserving collaborated optimization. To appropriately distribute interests within the operation, a quotation mechanism inspired by the Vickrey–Clarke–Groves (VCG) mechanism is introduced. Case studies verify that the proposed framework significantly reduces the total actual cost by 37.58% under a specific workload level. The cost reduction or net profit for each participating enterprise is also guaranteed, validating the efficacy and practicality of the proposed framework. • Allocates workload to cross-enterprise geo-distributed data centers. • Shares idle computing resources, improving flexibility in workload allocation and reducing costs. • Proposes an algorithm that combines ADMM with homomorphic encryption to ensure privacy. • Introduces a VCG-based mechanism that fairly distributes interests and encourages enterprise participation. • Total cost is reduced by 37.58%, and each enterprise’s cost savings or profit is also confirmed.