In this talk, I will present the KPC-Toolbox, a MATLAB tool for fitting workload traces into Markovian Arrival Processes (MAPs). MAPs are a general class of Markov-modulated processes used for fitting real workload traces with time-varying characteristics, e.g., for approximating workloads with short-range or long-range dependent behavior. After introducing the main chacteristics of MAPs, I will build intuition on which trace descriptors are most important to be fitted in MAPs for queueing model parameterization. Given that the MAP state space can be large, KPC-Toolbox focuses on first determining the order of the smallest space that can fit the trace well and then it derives the MAP that captures accurately the most essential features of the trace. Experimentation on temporal dependent traces illustrates the effectiveness of the KPC-Toolbox in fitting traces that are well-documented in the literature as very challenging to fit.

Giuliano Casale received the MEE and PhD degrees in computer engineering from the Politecnico di Milano, Milan, Italy, in 2002 and 2006, respectively. He is currently a researcher at SAP Research, CEC Belfast. From January 2007 he was postdoctoral research associate at the College of William and Mary, Williamsburg, Virginia, where he studied the performance impact of burstiness in systems. In Fall 2004 he was a visiting scholar at UCLA studying bounds for queueing networks. His research interests include performance evaluation, modeling, capacity planning, and simulation. He is a member of the ACM, IEEE, and IEEE Computer Society.