Empirically learned characteristics of a business process, like
branching probabilities and average execution durations of activities, are usually
used for process simulation and the creation of scheduling scenarios as a basis
for process optimization. This information can also be utilized to calculate a
probabilistic timed process model, based on the process structure. It yields time
histograms which represent the probability for expected process executionduration
and activity execution-intervals. In order to cope with the interoperability
aspect of business processes these workflow-based concepts must be
mapped on composite web service (CWS) environments. Slow autonomous web
services, invoked by a CWS, can have a disastrous impact on the overall process
response time. Thus techniques are needed predict the process duration based
on the anticipated response time of participating web services, which enables us
too avoid these services or to optimize them for faster execution.