Workload-Driven Database Optimization for Cloud Applications

The performance of modern data-intensive applications is closely related to the speed of data access. However, a physical database optimization by design is often infeasible, due to the presence of large databases and time-varying workloads. In this paper we introduce a novel methodology for physical database optimization which allows for a quick and dynamic selection of indexes through the analysis of database logs. The application of the technique to cloud applications, which use a pay-per-use model, results in immediate cost savings, due to the presence of elastic resources. In order to demonstrate the effectiveness of the approach, we present the case study Nuvola, a SaaS multitenant application for schools that is characterized by heavy workloads. Experimental results show that the proposed technique leads to a 52.1% reduction of query execution time for a given workload. A comparative analysis of database performance before and after the optimization is also performed through a M/M/1 queue model and the results are discussed.