Sharing is Caring A Decision Support Model for Multi-Tenant Architectures

Business software is increasingly moving from a traditional on-premises deployment model to a Software as a Service deployment model. In a Software as a Service deployment model, the possession and ownership of the software application is separated from its use. The software is hosted by a Software as a Service provider, relieving the customer organization from the responsibility for supporting the software, and purchasing and maintaining server hardware for it. The service provider can achieve substantial cost savings by applying economies of scale. This involves a system where customers share services, databases or resources and is known as multi-tenancy. The option to enable multi-tenancy is not binary. There exist various multi-tenant architectures, because it can be applied at different levels in the architecture. Also, multi-tenancy is not necessarily beneficial, certain situations require a more single-tenant approach. The appropriate level of resource sharing is crucial for a software provider, because it defines an architectural decision. However, there is insufficient knowledge and understanding to determine the most suitable multi-tenant architecture for the software application of a specific Software as a Service provider. This research focuses on the development of a Multi-Tenant Architecture Selection Model to assist service providers with this architectural decision problem. First by means of a structured literature study a set of twelve multi-tenant architectures is identified. These multitenant architectures describe which resources in an application’s system are shared among tenants, discriminating between the application and database layer. With the same literature study a list of twenty two decision criteria, representing factors that influence the decision, is identified. They are based on consequences, drawbacks and benefits, considerations and requirements related to multi-tenancy. The multi-tenant architectures and decision criteria are then evaluated by domain experts. The Analytic Hierarchy Process is selected as the decision making method, based on the complexity of the decision problem, the lack of quantitative data, and the importance of weighing the decision criteria. After this selection, all multi-tenant architectures are ranked on each decision criterion, using domain experts. This results in a decision matrix showing the performance score of each architecture with respect to each criterion. This matrix can then be used by Software as a service providers performing the analytic hierarchy process. The set of multi-tenant architectures, the list of decision criteria, the decision matrix and the final Multi-Tenant Architecture Selection Model are the key deliverables of this research and support architects in choosing the most suitable architectural pattern. This research is the first step in helping architects of Software as a Service providers make better architectural decisions, saving them time, effort and potential problems in the future.