School of Electrical Engineering and Computer Science, University of Central Florida

 Performability Module This module consists of the following four parts: Part 1: Fundamental Performance Evaluation Concepts This part gives the basic definition of a “system” and introduces the principal measures for system dependability and system performance. The two important concepts of Quality of Service (QoS) and Class of Service (CoS) are explained and illustrated by examples. Part 1 is concluded by giving a classification of the techniques used in performance evaluation and explaining the steps involved in conducting a modeling study. Part 2: Basic Tools in Markov Chains This part covers the necessary analytical background needed for Part 3. The basic concepts of Markov chains and exponential distributions are presented. Necessary and sufficient conditions for reaching a steady state (equilibrium) are given. The balance equations of an example “birth and death” Markov process are derived and solved. Part 3: Introductory Queuing Theory This part starts by explaining what a queuing system with a single server is. The measurable quantities and the product form solution associated with this queuing system are given. Kendall’s notation (shown in Fig. 1) for describing simple queuing systems is introduced. Solution of the M/M/1 queue is presented, Little’s formula is derived, and the M/GI/1 queue is analyzed. Examples of using the elementary M/M/1 and M/GI/1 queues in the performance evaluation of computer disk facilities are given. Part 4: Queuing Network Modeling and Performance Evaluation Part 4 begins by giving examples of representing computer systems as networks of queues that may be evaluated analytically. Jackson networks, BCMP queues and Gelenbe networks are presented as exact analytical solutions techniques that can be used to evaluate system performance in a variety of applications. A detailed example of a computer system with virtual memory is given and its solution as a closed Jackson network is presented. The non-analytical approach of discreet event simulation (DES) is subsequently covered along with the important issues of DES time management, event generation and synchronization, and DES statistical validation. Part 4 is concluded with a coverage of approximate solutions techniques. The approach of diffusion approximations is introduced and is applied to general networks of queues with one customer class. The approach is then used to solve a practical application involving packet switching networks. Finally the approach of fixed-point iterations is covered and is used to solve a practical application involving a distributed shared memory system.

 University of Central Florida (UCF) National Science Foundation (NSF)