Tenant Oriented Lock Concurrency Control in the Shared Storage Multi-tenant Database

SaaS provides multi-tenant technology allowing database to share storage but coming with irregular load, untraceable transaction characteristics and so on. Lock technology is important to ensure transaction concurrency control, but if multi-tenant database uses traditional granularity database locks, it will cause numerous problems because of the untraceable transaction characteristics and other reasons. To satisfy requirements in multi-tenant database, we propose TOL mechanism. It provides a lock of one granularity between the upper coarse-grained granularity and the lower fine-grained granularity for each tenant. Then we establish the transaction model of multi-tenant database to give comprehensive analysis of the transaction conflict, deadlock, and throughput and lock number of TOL. Meanwhile, Cluster and Diffusion methods are proposed to support the granularity adjustment for a single tenant. Combined with the analysis in the transaction model, a lock granularity adjustment strategy is built. It focuses on evaluating an approximate suitable granularity for one tenant between upper coarse-grained and lower fine-grained. In addition, we give experiments indicating that the performance of TOL is excellent and the lock granularity adjustment algorithm is feasible.