Performance Comparison of Middleware Threading Strategies

The spectacular growth of E-business applications on the Internet has boosted the development of middleware technology. Middleware is software that manages interactions between applications distributed across a heterogeneous computing environment. In the competitive E-business market the ability to deliver a high and predictable performance of E-business applications is crucial to avoid customer churn, and thus loss of revenue. This raises the need for service providers to be able to predict and control performance. The performance of middleware-based applications depends strongly on the choice of the so-called threading strategy, describing how the middleware layer handles competing method invocation requests. The goal of this paper is to provide an understanding of the impact of threading strategies on the performance of middleware- based applications. To this end, we (1) develop new quantitative models for the performance of middleware under different threading strategies, (2) perform extensive test lab experiments to compare the performance under different threading strategies, and (3) explain the experimental results by relating them to the quantitative models. As such, this paper provides new and fundamental insight in the impact of threading strategies on the performance of E-business applications.