Internally funded project
Start date : 01.11.2004
End date : 31.12.2005
In cooperation with Chair 4 (Distributed Systems and Operating Systems), the queueing network analysis tool WinPEPSY-QNS (Windows Performance Evaluation and Prediction System for Queuing Networks) is being developed. The analytical methods used include, among others, the mean value analysis (MVA) and the method of Marie. For validation and analysis of non-product form networks, a simulation component will be integrated into WinPEPSY-QNS. In addition, the Jackson method is being used to analyze open product form networks and a decomposition method for open non-product form networks. Of particular note is the possibility to present the results in a tabular or graphic form in a very clear manner and to easily carry out extensive experiments for a given queueing network.
The analysis methods for queueing networks with general distributions developed in the project Analysis Methods for Non-Markovian Models are to be integrated into WinPEPSY-QNS. They allow the approximative analysis of non-product form networks based on the method of supplementary variables or modeling by phase-type distributions and avoid the problem of state space explosion. Therefore, queueing systems with any distribution of service times and inter-arrival times can be analyzed. Emphasis should be placed on heavy-tailed distributions and the deterministic distribution. Thus, it should be possible to investigate the influence of methods and systems for the protection of privacy, authentication, guarantee of integrity as well as the anonymization on the performance (throughput and delay) through queueing systems.