Simulation of a Message Switched Network

Rockwell International Science Center; Thousand Oaks, CA

A message switched network may be regarded as a collection of physical circuits interconnected by queuing/routing points called nodes. A message is defined here to be a logical unit of information from the standpoint of the user. It consists of a finite length bit string. Messages enter the network at the nodes, which store and forward these messages as they are routed to their destinations. Thus, an individual message may pass through several nodes which buffer or queue the messages and attempt to optimize the routing from node to node to increase the total throughput of the network. Routing algorithms can range from fixed protocols, to simple random selection, to those that adapt to changing traffic and topology.

A 5-node store-and-forward message switched network was the basis for this discrete event message level simulation model. Messages are created at the nodes, work their way through the network, queuing at each of the nodes and links, and are destroyed at their destination. The network consists of three types of components: message generators, links, and nodes. All within each type are functionally identical and are modeled as asynchronous concurrent "processes" in SIMSCRIPT II.5.

The conventional analytic approach to the study of message switched networks rests upon simplifying assumptions regarding the statistics of message input, node service time and routing (switching) policies, severely limiting the class of networks that can be studied in this manner. A message level simulation, on the other hand, can be used to relax many of those restrictive assumptions. Furthermore, the internal logic of the nodes in the network can be simulated to any level of detail. Random processes within the model can be easily varied in order to investigate the sensitivity of the results to the underlying statistical assumptions.

Benefits of SIMSCRIPT II.5: The programming effort needed to investigate differing network topologies and traffic statistics using SIMSCRIPT II.5 can be substantially reduced over that needed using a non-problem oriented language. The simulation approach can readily provide estimates of not only the equilibrium distributions of network performance measures, but of the long and short-range dynamic characteristics of these parameters.

Customer Quote: "SIMSCRIPT II.5 contains language elements and software mechanisms designed for handling the kinds of concepts encountered in the modeling and simulation of communications networks."