Advanced monitoring and smart auto-scaling of NoSQL systems

Recent years have shown that RDBMS systems do not always meet the performance and scalability requirements of today's applications. Horizontal scalability is hindered by the ACID properties and the normalized data model these systems use. For this reason, a whole new range of database systems (NoSQL systems) has emerged. This paper focusses on eventual consistent storage systems, which have a certain inconsistency window after an update. Within this window different replicas contain a different version of a certain data item. While RDBMS systems provide strong transactional semantics, this is not the case for eventual consistent storage systems. The level of consistency is often configurable, but figuring out the optimal configuration is not a trivial task. Next to that, recent research has shown that the size of the inconsistency window can change over time, considering a fixed configuration. In this Phd research we envision a solution where all consistency-related parameters are managed by an SLA-driven autonomous system. Continuously monitoring the size of the inconsistency window allows dynamic reconfiguration and re-provisioning of the database cluster to keep the inconsistency window under a certain limit. As such, more guarantees can be provided to the application programmer.