Performance of middleware based architectures: a quantitative approach
Most of today's E-business applications on the Internet are built upon middleware-based architectures. For service providers offering these applications performance is essential: less-than-acceptable performance levels may lead to customer churn, and thus loss of revenue, and as such directly affect the company's competitive edge. This raises the critical need for service providers to be able to predict and control performance. In this paper we demonstrate the usefulness of a quantitative modeling approach to analyze and predict the performance of middleware-based applications. To this end, we develop a quantitative performance model of middleware architectures based on CORBA, the de-facto standard for object middleware. A particular feature of the model is that it explicitly takes into account priority mechanisms that handle the access to the processors among the different threads. To validate the model we have compared performance predictions from simulation runs with results from lab experiments for a variety of parameter settings. The results show that (1) the inclusion of priority mechanisms in the model leads to a significant improvement of the accuracy of the performance predictions based on the model, and (2) a quantitative modeling approach to assess and predict the performance of middleware-based applications is very promising.
|, , , , , ,|
|30th International Computer Measurement Group Conference, CMG 2004|
Harkema, M, Gijsen, B.M.M, & van der Mei, R.D. (2004). Performance of middleware based architectures: a quantitative approach. In Proceedings of the 30th International Computer Measurement Group Conference, CMG 2004.