Versioning in Main-Memory Database Systems From MusaeusDB to TardisDB

As relational database systems do not support collaborative dataset editing, online lexicons-such as Wikipedia's MediaWiki-build their own version control above the database system to allow constraint-preserving version checkouts or commits involving multiple tables. To eliminate the need for purpose-specific solutions, we propose adding version control as a layer on top of the database system or integrating versioning in the database system's core. This paper presents the first two architectures for versioning an entire state of a database system with respect to references among multiple relations. We design the prototype MusaeusDB as a solution for existing database systems, either as an external tool or as an extended SQL interface. The prototype TardisDB-an extended main-memory database system-reuses multi-version concurrency control for in-place updates while keeping older versions accessible. For performance tests on different storage layouts, we create-based on Wikipedia's page history-the TardisBenchmark. Our results show that it is indeed feasible to reduce wasted space while still ensuring constant retrieval time. Also, extending a main-memory database system's multi-version concurrency control has no negative impact on the transactional throughput. For further research on database versioning, we offer a flexibly sized benchmark with time evolving, text-based datasets and compression techniques.