Abstract

Nowadays, the demand for robust, reliable, and dependable telecommunication systems is higher than ever. End users expect services to be delivered with minimal to no interruption especially in cases where the effect of service outage can have catastrophic consequences such as loss of human lives and monetary losses. Examples of such applications include air traffic control and navigation systems, or systems that perform money transfer transactions such as VISA. Systems are considered highly available if they are up and running 99.999% of the time. One solution to sustain such availability is for such systems to be deployed on specific middleware that allow the redundancy of the system components to ensure the availability of services they provide. However, most existing platforms are proprietary and platform dependent. The goal of the Service Availability Forum (SAF) is to develop open specifications that aim to standardize the interface between the applications and the middleware from one side and the middleware and the underlying hardware from the other side. SAF specifications have also been developed to allow highly available applications to be built using commercial off-the-shelf components. A key component of SAF is the Availability Management Framework (AMF), which is the middleware part responsible for managing the redundant resources of applications and therefore enables high availability. AMF, however, requires a certain organization and groupings of those components known as an AMF configuration. Creating AMF configurations manually tends to be very difficult, error prone and sometimes impossible when the number of components forming the application and the cluster hosting the application is considerably high, which is the case for most real-world telecommunication systems. In this thesis, we devise a solution for automatically generating AMF compliant configurations for applications. The proposed solution encompasses two techniques that vary depending on the way AMF entities are handled. We have implemented both approaches and applied one of them to a case study to demonstrate the applicability of our solution