Using Queuing Models for Large System Migration Scenarios – An Industrial Case Study with IBM System z

Abstract

Large IT organizations exchange their computer infrastructure on a regular time basis. When planning such an environment exchange, it is required to explicitly consider the impact on the Quality-of-Service of the applications to avoid violations of Service Level Agreements. In current practice, however, using explicit performance models for such estimations is frequently avoided due to scepticism towards their practical usability and benefits for complex environments. In this paper, we present a real-world case study to demonstrate that a queuing model-based approach can be effectively used to predict performance impact when migrating to a new environment in an industrial context. We first present a general modeling methodology and explain how we apply it for system migration scenarios. Then, we present a real-world industrial case study and show how the performance models can be used. The migration is planned for a System z environment running a large scale banking application. Finally, we validate the performance models after the system has been migrated, evaluate the prediction accuracy, and discuss possible limitations. Overall, the measurements show very high agreement with the prediction results.