[show abstract][hide abstract] ABSTRACT: Recent studies show the existence of a demand for a connectionless broadband service. In order to cope with this demand, a connectionless protocol for the B-ISDN needs to be designed. Such a protocol should make use of ATM and the ATM Adaptation Layer. It needs to specify destination and bandwidth of connections to the ATM network without advance knowledge of the traffic that has to be transferred over these connections. A possible mechanism which can cope with this problem, the `On-demand Connection with Delayed Release' (OCDR) mechanism, is described. Its efficient operation is based on the assumption that there exists a certain correlation between subsequently arriving CL packets. Two different arrival processes are used to evaluate the performance of the OCDR mechanism: a Poisson arrival process, and a Markov Modulated Poisson Process (MMPP) which models a bursty traffic source. Markov models of the OCDR mechanism have been constructed for both arrival processes. For the model with Poisson arrivals, a closed form solution is presented. The model with MMPP arrivals is solved numerically.
[show abstract][hide abstract] ABSTRACT: In this paper a new technique, called Fault Injection Simulation (FIS), is presented that is suited for deriving results for steady-state measures in discrete event dynamic systems which are strongly influenced by rarely occurring events. FIS is based on decomposition of the observations in those that are affected and those that are not affected by these events. If methods are available FIS can be used as a (partly) analytical technique, else as a pure fast simulation technique. It is shown that under intuitively appealing assumptions FIS gives an unbiased estimator and a variance reduction. Furthermore it is shown how to adjust FIS during the simulation to obtain maximum variance reduction. FIS is discussed in the context of rarely occurring failures in networks for realtime applications. Comparisons with other fast simulation techniques are made and results of FIS for an M/M/1-queue with rarely occurring service breakdowns are included.
[show abstract][hide abstract] ABSTRACT: Over the last two decades a considerable amount of effort has been put in the development and application of matrix geometric techniques for the analysis of queueing systems of which the (embedded) Markovchain exhibits a regular structure. Most of this work however has been presented in either a mathematical context or in a purely application oriented context.
[show abstract][hide abstract] ABSTRACT: In this paper we discuss a unified approach towards model-based quantitative evaluation of both computer systems and communication systems. In the area of fault-tolerant computer systems, dependability evaluation has been recognized as being a topic of importance, both to judge a system on its merits and to provide trust in the actual dependability of the system. In communication systems, the need for identifying and evaluating quality of service parameters is becoming more and more apparent because of increasing demands on for instance speed and availability. In this paper we construct a framework, the so-called performability evaluation framework, within which the quantitative evaluation of both types of systems can be discussed. The framework closely resembles the performability framework introduced by Meyer . We present a general view, a system view and a modelling view on performability evaluation, leading to a framework which naturally fits known measure definitions, modelling methods and solution techniques. We will especially discuss the importance of evaluating useful performability measures. In this respect we will distinguish between system measures and task measures. Task measures directly relate to evaluation of the service, and we will argue that in performability evaluation task measures should be evaluated. We relate the performability evaluation framework with known concepts in both the areas of computer and communication systems.
[show abstract][hide abstract] ABSTRACT: Over the last two decades a considerable amount of effort has been put in the development and application of matrix geometric techniques for the analysis of queueing systems of which the (embedded) Markov chain exhibits a regular structure. Most of this work however has been presented in either a mathematical context or in a purely application-orientedcontext.
[show abstract][hide abstract] ABSTRACT: Introduction. Stochastic Petri net-based Markov modeling is a potentially very powerful and generic approach for evaluating the performance and dependability of many dierent systems, such as computer systems, communication networks, manufacturing systems, etc. As a consequence of their general applicability, SPN-based Markov models form the basic solution approach for several software packages that have been developed for the analytic solution of performance and dependability models, e.g., [1, 8]. In these tools, stochastic Petri nets are used to conveniently specify complicated models, after which an automatic mapping can be carried out to an underlying Markov reward model. Subsequently, this Markov reward model is solved by specialized solution algorithms, appropriately selected for the measure of interest. State-space explosion. One of the major aspects that hampers the use of SPN-based Markov models for the analytic solution of performance and dependabilit