The IEEE token bus-A performance bound on GM MAP - Industrial Electronics, IEEE Transactions on

All versions of the General Motors Manufacturing Automation Protocol (GM MAP) specify that MAP’S lower layer services are to be provided by the IEEE 802.4 token bus. An understanding of 802.4 and its performance aspects is therefore prerequisite to predicting the performance of MAP. We show how total bus capacity is divided among data throughput, token traffic, and propagation delays. We show the relative contributions of access delay and queueing delay to total message delivery time. The effects on message delivery times of message size and the number of active stations are also reported. As the token cycle time increases beyond the Target-Rotation-Time for each of the asynchronous access-classes, service to the lower priority classes is curtailed; we present a formula which can be used to identify the offered load at which the transition from normal to curtailed service begins. I . THE IEEE 802.4 TOKEN Bus HE GENERAL MOTORS Manufacturing Automation T Protocol (GM MAP) [l] specifies that its lower layer services be provided by IEEE 802.4, a standard [2] for local area networks (LAN’s) that uses an explicit token passing scheme to control access on a bus topology network. The standard specifies the actions and services of the Medium Access Controller (MAC), the interface between the data link and physical layers of the IS0 OS1 model [3]. Although its physical topology is a bus, 802.4 creates a logical ring of active stations through the token passing process. 802.4 offers four priority classes for message transmissions and, for the highest priority messages, provides bounded token cycle times. The individual stations in an 802.4 network are free to leave and join the token passing ring as dictated by their traffic or station management decisions. The standard describes a robust protocol able to withstand the loss or duplication of tokens and the failure or disconnection of stations. Since the performance of GM MAP is bounded by the performance of the underlying token bus, we have studied that protocol via simulation to evaluate bus utilization, token cycle time, and message delay. We investigated the performance effect of varying message size and number of active stations. For the asynchronous access-classes, we show how the setting of the Target-Rotation-Times affects message delivery delay. Manuscript received February 12, 1987; revised July 24, 1987. This work was supported in part by the Institute of Information Technology of the Virginia Center for Innovative Technology, Herndon, VA 22071. A. C. Weaver is with the Department of Computer Science, University of Virginia, Charlottesville, VA 22903. C. F. Summers is with the Corporation for Open Systems, McLean, VA 22 1024306. IEEE Log Number 8718151. 11. SIMULATION STUDIES A Pascal program that simulates the actions and state transitions of user configurable 802.4 networks was used to study the performance of 802.4 and the effects of varying important configuration parameters. Unless otherwise noted, all simulations used the same basic configuration: an errorfree 10-Mbps bus, the smallest possible token, functionally infinite High-Priority-Token-Hold Time, or HPTHT (10 s), all traffic in the Synchronous access-class only, a logical ring of 64 stations with identical message creation rates and constant message sizes of 160 data bits (for a total of 256 bits in the frame), and all stations are always members of the token passing ring. Offered load was varied from 10 to 95 percent (increasing in steps of 5 percent per simulation) of the bus capacity by increasing the message creation rate h. For example, to achieve an offered load of 10 percent of the bus capacity with the above configuration