An automated approach for creating workload models from server log data

We present a tool-supported approach for creating workload models from historical web access log data. The resulting workload models are stochastic, represented as Probabilistic Timed Automata (PTA), and describe how users interact with the system. Such models allow one to analyze different user profiles and to mimic real user behavior as closely as possible when generating workload. We provide an experiment to validate the approach.