Availability Monitor for a Software Based System

Computer and communication systems are ubiquitous and are used extensively in safety critical, life critical, and finance critical applications. Due to the excessive cost of outages, downtime is not tolerated by the users. High availability applications are being offered by most vendors and are touted to have availability features such as hardware redundancy, software replication, automated detection, failover, hot swap and so on. However, quantitative validation of high availability is rarely provided. The purpose of this paper is to present an innovative method of monitoring and displaying, in real-time, the empirically observed availability of the system. Apart from presenting the novel idea and associated statistical methods, we also sketch the actual implementation of this idea in a project (a middleware appliance) at the WebSphere Institute of IBMRTP.