Search

Nowadays high availability and reliability are essential characteristics of any enterprise client-server architecture. This primarily requires a stable and fault-tolerant operation of the server side which can hardly be guaranteed without possibility to select an appropriate server depending on specific client needs. In this case fault-tolerance is provided by the client side. One of the main principles used for such approach is Recovery Block. Described in this paper is a new algorithm for increasing fault tolerance in client-server communication. It is based on Recovery Block method and allows ranking of available servers with their stability factor and making decision of removing unstable servers considering previous connection attempts.