Performance Evaluation of Distributed Computer-Communication Systems

Abstract

Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. The computer-communication field is rich with extremely challenging problems for the queueing theorist. In this paper, we describe a number of computercommunication applications in wide-area networks, packet radio networks and local area networks. All of these involve sophisticated queueing theoretic models which have led to advanced applications of existing theory and in some cases, to new methods in queueing theory. A significant component in many of these applications is that ot multi-access to a common resource (typically, a communication channel); we devote a significant portion of this paper to multi-access systems. In addition to identifying problems of interest to the queueing theorist, this paper summarizes the latest results for the mean response time of many current computer-communication applications1. INTRODUCTION The advent of the digital computer brought about a major rebirth in applied queueing theory. Indeed, since the early 1960's we have seen a sequence of applications in the information processing field which have challenged the capabilities of that theory and which have forced the theory to be extended in new directions. There is no reason to believe that thai challenge will subside for some time to come. The response to these challenges has produced some important extensions in queueing theory. For example, there has been a major (and largely successful) assault on queueing networks of various kinds. Some of the results are quite beautiful, and exhibit robustness in various directions (see (81]). We have also seen the development of some very important approximation techniques. Multi-access and broadcast communication problems have forced us to develop further analytical techniques. Many of the applications lead to coupled queues which involve two-dimensional queueing problems in the simplest cases, and higher dimensions in the usual cases: here too. new approaches in queueing theory have had to be developed (e.g.. the solution of RiemannHilbert problems). Due to the difficulty of many of these problems, we find 1.

Internet congestion control using the power metric: Keep the pipe just full, but no fuller

“…In [33] , the ZAP approximation was introduced as a family of response time functions to represent the performance of various systems of flow. Here we follow that approach and consider the following three-parameter expression for T ( ρ),…”

From computer science to service science: Queues with human customers and servers

“…In order to answer this question mathematically, a metric of system performance called power was proposed and studied. Although the notion of power had been introduced earlier by others [7,27], I believe it is Professor Kleinrock who exploited its significance intensively and extensively [15,16,17]. The power was later studied in detail Table 1: Trade-off of resource utilization and customer satisfaction in the M/M/m queue.…”

Applications of polling systems

“…This type of protocols naturally leads to the formulation of polling models with random routing [36,98]. A detailed comparison of various multi-access schemes, including random access schemes like ALOHA and CSMA, is given by Kleinrock [97].…”