Modelling techniques and tools for computer performance evaluation

Welcome to this special section of Performance Evaluation, presenting expanded versions of selected papers from the 13th International Conference on Modelling Techniques and Tools for Computer Performance Evaluation. The series of TOOLS conferences has provided a forum for the community of performance engineers with their diverse interests to interact with one another in a wide variety of theoretical and applied areas. TOOLS 2003 was the second conference in the TOOLS series to be held in the state of Illinois, USA. It was one of four component conferences that met together under the umbrella of the 2003 Illinois Multiconference on Measurement, Modelling, and Evaluation of Computer-Communication Systems at the University of Illinois at Urbana-Champaign. The other conferences held in conjunction with TOOLS 2003 were the 10th International Workshop on Petri Nets and Performance Models (PNPM 2003), the 2003 International Conference on the Numerical Solution of Markov Chains (NSMC 2003) and the Sixth International Workshop on Performability Modelling of Computer and Communication Systems (PMCCS-6). TOOLS 2003 had 37 regular submissions, of which 17 were selected as full papers. The proceedings of TOOLS 2003 have been published in the Lecture Notes in Computer Science series (vol. 2794, P. Kemper, W.H. Sanders (Eds.)) by Springer-Verlag, as has been the tradition in this conference series since 1994. In the current special issue, a selection of three papers from TOOLS 2003 is presented. These papers have been selected by the Program Committee, and have been extended and considerably improved by their authors to reach their current form. An additional review procedure and revision process took place, and final versions of the papers were delivered in Spring 2005. The first paper, entitled Closed-Form Solutions for Mapping General Distributions to Quasi-Minimal PH Distributions, by Takayuki Osogami and Mor Harchol-Balter (CMU), considers the approximation of general distributions as phase type distributions. In particular, the authors propose an algorithm for mapping a general distribution G to a phase-type distribution that matches the first three moments of the original distribution. The efficiency and accuracy of their approach result from a beneficial selection of a class of distributions. By combining Erlang-n and Coxian distributions, the authors obtain a distribution with a small number of parameters for which closed form solutions are obtained as well as a nice result on the minimality of the number of phases in the resulting distribution. In the second paper, entitled Correlation Bounds for Second-Order MAPs with Application to Queueing Network Decomposition, Armin Heindl (University of Erlangen-Nuremberg), Ken Mitchell, and Appie