Formal Modeling and Analysis of Cassandra in Maude

Distributed key-value stores are quickly becoming a key component of cloud computing systems. In order to improve read/write latency, distributed key-value stores offer weak notions of consistency to clients by using many complex design decisions. However, it is challenging to formally analyze consistency behaviors of such systems, both because there are few formal models, and because different consistency level combinations render understanding hard, particularly under communication latency. This paper presents for the first time a formal executable model in Maude of Cassandra, a popular key-value store. We formally models Cassandra’s main components and design strategies. We formally specify various consistency properties and model check them against our model under various communication latency and consistency combinations.