Design and Validation of Cloud Storage Systems Using Rewriting Logic

Most large modern-day applications (Facebook, Gmail, eBay, etc.) rely on widely distributed and replicated storage of data for scalability, availability, and disaster tolerance. Since maintaining high degrees of data consistency requires costly communication across distant sites, applications over such distributed and partially replicated data are complex artifacts that must carefully balance the required degrees of consistency and performance. In this paper I summarize work at the University of Illinois Assured Cloud Computing center on using rewriting logic and its associated Maude tool environment to formally model and analyze both the correctness and the performance of state-of-the-art distributed transaction systems designs, as well as on how to automatically obtain a correct-by-construction distributed implementation of a promising design.