Reasoning about block-based cloud storage systems via separation logic

Abstract

Owing to the massive growth in the storage demands of big data, cloud storage systems emerge at a historic moment, among which block-based cloud storage systems (BCSSs) are typical. Compared with traditional storage systems, BCSSs have many advantages, such as higher capacity, lower cost, and better scalability. BCSSs are also much more complex and error prone than traditional systems, and therefore, their reliability is a significant concern. Correctness is the most basic requirement for reliability. Therefore, it is of great significance to study BCSS program correctness verification to establish the theoretical basis for cloud storage systems. In this paper, we propose a novel framework based on separation logic (SL) to verify the correctness of BCSS management programs. We first construct a two-tier heap structure as the formal model of the storage architecture for BCSSs and present a modelling language based on this structure to describe the program execution. We then define assertion pairs to describe the properties of the two-tier structure. The two components of an assertion pair affect each other, producing several interesting properties. Finally, we propose a Hoare-style proof system to reason about BCSS programs and establish its soundness. We also demonstrate our proof system's uniqueness, usability, and scalability by verifying several real-world BCSS sample programs. In summary, this work presents a systematic approach that can be used to effectively solve BCSS program verification problems, and provides an SL-based logical modelling strategy, specifically to build its consistency with the actual behaviour of BCSSs.