Abstract

Information Technology service providers aim at attracting customers by providing services that meet a high level of quality. They should not only satisfy functional requirements, but also non-functional requirements. An important non-functional requirement is the level of service availability. Service Availability Forum (SAForum), a consortium of communications and computing companies, has developed a set of services and standard Application Programming Interfaces (APIs) to address the issue of high availability for the Commercial-off-the-shelf (COTS)-based systems and enable portability. Among the services standardized by the SAForum, the Availability Management Framework (AMF) has the important role of ensuring the high availability of an application and its services, by managing the redundant application components deployed on the cluster. To achieve this task, AMF requires a configuration that represents the logical organization of the application components and their services.
The design of AMF configurations is a complex and error prone task. Automation of the process is the first step towards improving the quality of such configurations. It also enables exploring different potential solutions for a given set of requirements. An automated approach to generate configurations for applications to deploy on top of the SAForum middleware has been proposed in the context of the MAGIC project. This approach, however, may generate several configurations among which some may not meet the required level of service availability. Therefore, the system designer needs to evaluate the generated configurations using an availability analysis tool to select an appropriate one for deployment. One may want to improve this process by targeting directly in the generation process, the configurations that can guarantee the requested level of service availability.
The objective of this thesis is to propose solutions to enhance this configuration generation process and generate configurations that can guarantee the required level of service availability without using advanced analysis tools. For this purpose, we propose configuration design patterns to improve the expected level of service availability and quantitative methods that eliminate some configurations that do not meet the availability requirement. The configuration design patterns improve the expected level of service availability by selecting the best configuration options. The methods estimate service availability for the different possible combinations of software components, which can provide the requested services, taking into account the properties of these components and the behavior of the SAForum middleware. As a proof of concept, we have embedded our proposed solutions into a prototype tool as an eclipse plug-in and validated our work with case studies.