Evaluation of Prioritization in Performance Models of DTP Systems

Abstract

Modern IT systems serve many different business processes on a shared infrastructure in parallel. The automatic request execution on the numerous interconnected components, hosted on heterogeneous hardware resources, is coordinated in distributed transaction processing (DTP) systems. While pre-defined quality-of-service metrics must be met, IT providers have to deal with a highly dynamic environment concerning workload structure and overall demand when provisioning their systems. Adaptive prioritization is a way to react to short-term demand variances. Performance models can be applied to predict the impacts of prioritization strategies on the overall performance of the system. In this paper we describe the workload characteristics and particularities of two real-world DTP systems and evaluate the effects of prioritization concerning supported overall load and resulting end-to-end performance measures.