Formal Modeling and Analysis of Google’s Megastore in Real-Time Maude

• Published in: Specification, Algebra, and Software Vol. 8373; pp. 494 - 519
• Main Authors: Grov, Jon, Ölveczky, Peter Csaba
• Format: Book Chapter
• Language: English
• Published: Germany Springer Berlin / Heidelberg 2014; Springer Berlin Heidelberg
• Series: Lecture Notes in Computer Science
• Subjects: Entity Group; Message Delay; Model Check; Partial Order Reduction; Statistical Model Check
• ISBN: 3642546234; 9783642546235
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-642-54624-2_25

Cloud systems need to replicate data to ensure scalability and high availability. To enable their use for applications where consistency of the data is important, cloud systems should provide transactions. Megastore, developed and widely applied at Google, is one of very few cloud data stores that provide transactions; i.e., both data replication, fault tolerance, and data consistency. However, the only publicly available description of Megastore is short and informal. To facilitate the widespread study, adoption, and further development of Megastore’s novel approach to transactions on replicated data, a much more detailed and precise description is needed. In this paper, we describe an executable formal model of Megastore in Real-Time Maude that we have developed. Our model is the result of many iterations resulting from correcting design flaws uncovered during Real-Time Maude analysis. We describe our model and explain how it can be simulated for QoS estimation and model checked to verify functional correctness.