G. Michael Schneider

An algorithm for the design of multilevel concentrator networks

This paper presents an algorithm for the design of hierarchical, multi-level concentrator networks. The algorithm, based on the technique called clustering, is a generalization of an algorithm reported by P. McGregor and D. Shen and which appeared in the recent literature. The modification described here relaxes two of the assumptions made in the original McGregor/Shen formulation: 1. a) Solutions may have an arbitrary number of intermediate levels, instead of exactly zero or one. 2. b) The characteristics of a concentrator (i.e., capacity, cost, or local processing capability) can vary as a function of their placement within the solution. These relaxed conditions lead to solution networks which can be lower in cost than those generated by the algorithm in its original formulation. This paper will present the algorithm and show some comparative results.

Methodology and experimental research in software engineering

There has been a very rapid increase in the use of psychological experimentation as a method for addressing problems in the area of software engineering. However, there is a total lack of understanding on how to use this technique effectively so that the results of the experiment will have both validity and wide applicability. This paper describes what we feel is a major new research area which aims at formalizing the use of controlled group experimentation in programming language design and software engineering. The paper describes the growth of the problem, the critical need for this new research area and finally categorizes the problems that must be addressed as part of the research into these problems. An annotated bibliography contains literature references (through 1978) to the use of controlled group experimentation in software engineering and related areas of Computer Science.

A modeling package for simulation of computer networks

The VANS (Value Added Network Simulator) system re presents a new approach to the simulation of distrib uted resource-sharing computer networks. It allows the network designer either to manipulate the struc ture of the protocols (sets of rules) that govern the handling of communications in different types of networks or to investigate the effects of changing the parameters of networks operating under a fixed set of protocols. This paper describes the VANS sys tem and its use in conducting a wide range of experi ments in network design. While a number of simula tion programs permit experimenting with network parameters, VANS is the first to permit both that type of experimentation and experimentation with the protocols that control the operation of a computer network.

On the complexity of measuring software complexity

The term software complexity (program quality, program complexity,...) has been used by software engineering researchers to denote the testability, maintainability, readability, and/or comprehensibility of a program. Curtis points out a common bond that unites all these concepts: a programs complexity determines how difficult it is for programmers to work with.

A modular approach to computer network simulation

Most computer network simulations have allowed only simple parametric analysis of a specific network. The higher-level problems of network design such as routing, error handling, and flow control, have not been part of the user/model interface and are changed only with great difficulty. This communication describes a structural simulation model which allows a user to access, modify and experiment with the actual logical design of computer networks. This allows the model user to get specific quantitative information on the efficiency of alternate protocols for a particular environment and leads to meaningful network design decisions.

Yet another survey of graduate programs in computer science

This paper performs a quantitative ranking of the graduate programs in computer science at major universities in the United States and Canada. The rankings are based on the number of journal articles which have appeared in 12 major journals over the past two years. The paper also includes a ranking of major corporate research centers and foreign universities. The paper concludes with suggestions about how, in the future, such surveys should be conducted under the auspices of the professional organizations of computer science, e.g., ACM, SIGCSE.