An Empirical Case-Study of a Central-Server-Model on System Performance

摘要

This paper investigates from a practical point of view the applicability and the limits of a simple Central-Server-Model for performance modeling. To model the prolific webserver at the University of Applied Sciences in Kempten the study uses a Central-Server-Queuing-Theory- Model consisting of a single CPU server/queue and a single Disk server/queue. First the webserver is monitored and the model is calibrated using real data. Then performance is predicted using the model and the outcomes are compared to reality. In addition the model is validated analyzing the calculated and the measured response times with the aid of replaying a sequence of requests. A key point the paper deals with, is the question which and how much information gets lost using such a simple model.