Data consistency in multi-cloud storage systems with passive servers and non-communicating clients

Abstract

Multi-cloud storage systems are becoming more popular due to the ever-expanding amount of consumer data. This growth is accompanied by increasing concerns regarding security, privacy, and reliability of cloud storage solutions. Ensuring data consistency in such systems is especially challenging due to their architecture and characteristics. Specifically, passive cloud storage cannot run coordination software, and clients cannot communicate directly to reach consensus. Furthermore, the atomicity of operations is not always guaranteed by the clouds' public APIs. In this paper, we formally define data consistency in multi-cloud storage systems, identify how they can be violated, and introduce a new method that provably maintains the data consistency in these systems. The implementation and experiments showthat the proposed method can maintain data consistency with a certain delay in data uploading and that it is scalable with respect to the number of used clouds as well as the number of users. Integrating this method into multi-cloud storage systems will enhance their usability and reliability.