Automatic generation of performance models for distributed application systems

Organizations have become increasingly dependent on computing systems to achieve their business goals. The performance of these systems in terms of response times and cost has a major impact on their effectiveness. To achieve openness and scalability, these systems have begun to rely on distributed environment technologies such as the Open Software Foundation’s Distributed Computing Environment (DCE) and the Object Management Group’s Common Object Request Broker Architecture (CORBA). The resulting systems are complex and have performance behaviour that can defy intuition. Predictive performance models and performance evaluation techniques are needed to help manage their behaviour. Predictive models allow the timely evaluation of the performance impact of many different application loads and alternative configurations. Unfortunately model building by hand can be timeconsuming and error-prone. This paper describes techniques developed to help automate the construction of performance models for operational distributed application systems. It describes the information that is needed to automate model building, the infrastructure that supports the model building, and the methods used to obtain or estimate the required information. An example is given to illustrate the model-building process and the evaluation of several "what-if" scenarios for management that can be answered using