Autonomic Algorithms For Management And Evolution Of SaaS Systems

SaaS(Software as a Service) is a software delivered through the internet. SaaS software model is opposed to the traditional software model, which is purchased and installed on personal computers. SaaS is growing exponentially and numerous enterprises of all sizes can generate major performance improvements and cost savings through its application. This kind of application is a far more attractive economic model than the perpetual license model. The goal of the autonomic management is to develop applications and systems to manage themselves based on high perspective guidance from humans. There have been many chanllenges to research and development of autonmic algorithms for the QoS improvement and evolution of the SaaS applicaionts.In this thesis, the meta-model of SaaS applications is proposed for autonomic management and evolution. The meta-model addresses multiple layered structures of SaaS applications, from bottom to up, containing graphical databases, services, business processes, and user interfaces. The associations and dependencies of the layered constituents are described from different perspectives of providers, tenants, and users for SaaS applications.In order to enable SaaS applications dynamically adaptive to the changes of requirements from the SaaS providers, tenants and users, the hierarchy applicability requirements for the services and the QoS management requirements of the SaaS application are formally defined and described by using the formal languages of Probabilistic Computation Tree Logic (PCTL) and Continuous Stochastic Logic (CSL). And the Discrete SaaS User-Tenant-Provider Model (DSUTP) and Continuous SaaS User-Tenant-Provider Model (CSUTP) are proposed for QoS analysis and evaluation of SaaS applications.On the basis of the meta-model and models of SaaS applications, Autonomic Algorithm for QoS management of SaaS Applications (AAQS) is presented to improve the QoS of SaaS applications. The experimental verification and systematic analysis are conducted to evaluate the ability of dynamic detection and prediction of QoS violation in the SaaS application.Autonomic Algorithm for Evolution SaaS Application (AAES) is proposed to enhance the evolution capability of SaaS applications. The meta-heuristic (MH) and SaaS Case Based Reasoning (SaaS CBR) methods are developed for the selection of a SaaS application based on the best practices cases. The experimental results shows the AAES algorithm can improves the performance of SaaS application by selecting suitable cases for the SaaS providers, tenants, and users.Finally, by leveraging SaaS QoS management and evolution capability we proposed the Autonomic Algorithm for SaaS System Management (AASS) for autonomic management and evolution of the SaaS systems. The SaaSEHR, which is a SaaS based EHR system, is developed as the case study of the autonomic SaaS system. The prior and posteriori knowledge is used to enable SaaSEHR autonomic evolution. The experimental results show that AASS can promote the self-optimization and self-healing abilities of SaaS system by using suitable candidate services for SaaS tenants and users respectively. And the SaaS adaptability to changes of the environment and requirements are enhanced accordingly. Moreover, the performance of the SaaS system and the ability to control the SaaS services can be improved by sharing the workflow among SaaS tenants and users.