Comparative Analysis of Genetic Algorithm and XML Filtering Technique for Multi-Tenant SaaS Configuration Management

Abstract

The use of Software-as-a-Service (SaaS) cloud model is increasing rapidly. Among different SaaS deployment models, Single Instance Multi-Tenant (SIMT) has multiple advantages, e.g., high scalability and lower cost, but it also has practical concerns for management, as tenants may have distinct requirements or even conflicting preferences, and the number of features and tenants may be large. Offering the best user satisfaction, dealing with tenants' preferences and operational environment changes is crucial but challenging for effective SaaS management. Genetic algorithms (GAs) have been applied to the optimization problem in many fields. Further a GA-based approach, GAFES, specifically adapted to constraint-based feature selection optimization in Software Product Line (SPL) for multi-tenant SaaS has been reported. On the other hand, an XML filter technique, Yfilter, has been applied to various problem domains and recently to SIMT for SaaS. The aim of this paper is to experimentally evaluate the performance of GAFES and Yfilter for SaaS configuration with recourse constraints. We have conducted experiments to evaluate the two very different multi-tenant SaaS configuration techniques. The results show that both approaches can meet the user satisfaction, but Yfilter has much higher performance compared to GAFES.