Optimizing Replica Exchange Strategy for Load Balancing in Multienant Databases

Resource sharing in multitenant databases is a challenge issue. The phenomenon can adversely affect a tenant’s performance due to contending for shared resources among other tenants, and may cause a performance crisis. In this paper, a performance crisis is mitigated by a dynamic load balancing mechanism, which based on exchanging the roles between tenants’ primary replicas and secondary replicas. The mechanism is composed of two parts: firstly, to balance resource utilization across servers, queries are dynamically allocated according to resource consumption; secondly, Improved Simulated Annealing Algorithm is developed to identify an optimal subset of tenants on overloaded servers, and for each tenant in the set, a suitable secondary replica can be selected to exchange roles with its primary replica to mitigate a crisis. Experimental results show significant reduction of service level objective violations compared to migration-based load balancing method and no load balancing method, respectively.