Applying Stochastic Metaheuristics to the Problem of Data Management in a Multi-Tenant Database Cluster

摘要

A multi-tenant database cluster is a concept of a data-storage subsystem for cloud applications with the multi-tenant architecture. The cluster is a set of relational database servers with the single entry point, combined into one unit with a cluster controller. This system is aimed to be used by applications developed according to Software as a Service (SaaS) paradigm and allows to place tenants at database servers so that it may provide their isolation, data backup and the most effective usage of available computational power. One of the most important problems on such a system is an effective distribution of data between servers, which affects the degree of individual cluster nodes load and fault-tolerance. This paper considers the data-management approach based on the usage of a load-balancing quality measure function. This function is used during the initial placement of new tenants and also during placement optimization steps. Standard schemes of metaheuristic optimization such as simulated annealing and tabu search are used to find a better tenant placement.