Testing of transactional services in NoSQL key-value databases

Transactional services guarantee the consistency of shared data during the concurrent execution of multiple applications. They have been used in various domains ranging from classical databases through to service-oriented computing systems to NoSQL databases and cloud. Though transactional services aim to ensure data consistency, NoSQL databases prioritize efficiency/availability over data consistency. In order to address these issues various transaction models and protocols have been proposed in the literature. However, testing of transactions in NoSQL database has not been addressed. In this paper, we investigate into the testing of transactional services in NoSQL databases in order to test and analyse the data consistency by taking into account the characteristics of NoSQL databases such as efficiency, velocity, etc. Accordingly, we develop a framework for testing transactional services in NoSQL databases. The novelty and contributions are that we develop a context-aware transactional model that takes into account contextual requirements of NoSQL clients and the system level setting in relation to the data consistency. This can assist NoSQL application developers in choosing between transactional and non-transactional services based on their requirements of the level of data consistency. The framework also provides ways to analyse the impact of the big data requirements and characteristics (e.g., velocity, efficiency) on the data consistency of NoSQL databases. The evaluation and testing are carried out using a widely used NoSQL key/value database, Riak, and a real (open) and big data from the Council of London for public transportation of the London bus services.