Guest Editors' Introduction to the Special Issue on Quantitative Evaluation of Computer Systems

Abstract

OFTWARE systems are expected to meet a multitude of quantitative constraints, such as timely response, performability, and efficient use of energy and other resources. Quantitative evaluation of systems refers to a set of techniques which can be applied to analyze such systems and obtain a range of quantitative characteristics of their performance, for example, the probability of satisfying a request within a given time or the appropriate number of replicas in a quorum air traffic control system. In this special issue, we publish a diverse range of novel research work which contributes to the development of quantitative evaluation of computer systems from a number of perspectives. These include theoretical foundations, models and modeling languages, analysis methods, simulation, verification, and software tools. Queueing networks have long been used for performance analysis of computer systems. This issue includes two papers related to queueing networks. In the first, “Enhanced Modeling and Solution of Layered Queueing Networks,” the layered queueing networks modeling formalism is presented. This modeling language allows a structured, hierarchical queueing network to be built up representing the resource usage and contention within a system. Analysis techniques which make use of the hierarchical structure allow complex models to be analyzed to address questions about the performance and dependability of the system. The second queueing paper, “CoMoM: Efficient Class-Oriented Evaluation of Multiclass Performance Models,” describes an improved solution technique for multiclass queueing network models. For more than 25 years, mean value analysis (MVA) has been one of the most frequently applied analysis methods of these models. This paper proposes an efficient rearrangement of recursive computational steps (referred to as the CoMoM method) such that it minimizes the number of recursive steps needed to solve the model. The CoMoM method makes possible the approximate analysis of much larger queueing network models, with tens of classes and thousands of nodes. Another well-established quantitative modeling technique is based on timed and stochastic extensions of Petri nets. In “State-Density Functions over DBM Domains in the Analysis of Non-Markovian Models,” a solution technique for such a Petri net extension is considered. The analysis of stochastic Petri nets with exponentially distributed firing times is well-developed and widely applied in software systemsmodeling.Incontrast,thecaseofstochasticPetrinets