Using Analytic Models for Predicting Middleware performance

The client-server paradigm is very popular in building distributed computing applications. Heterogeneity is natural in client-server systems, where components implemented using different technologies must interact and collaborate with each other. Inter-operability is provided through middleware on such a heterogeneous system. Common Object Request Broker Architecture (CORBA) is a standard for middleware-based distributed object computing systems. This paper focuses on analytic modeling of middleware-based systems that can be used in the performance engineering of CORBA-based client server applications. The paper applies the Layered Queuing Network model to two types of middleware architectures: handle-driven and forwarding. The model outputs are compared with measured values to determine the accuracy of the modeling technique