Queuing models for computing and communication performance in distributed manufacturing control systems

A real-time scheduling algorithm called distributed arrival time control (DATC) has recently been developed for real-time scheduling of manufacturing systems. This algorithm uses feedback from distributed simulation to iteratively adjust schedules in real time to improve global performance. Therefore, the overall performance of DATC depends on the quality of service (QoS) provided by the computing and communication system. There is a need to develop effective performance models for systematically designing the computing and communication architecture in such applications. In this work we formulate an analytical model based on queuing theory to predict the network latency for a given workload and computing/communication architecture. Experimental data from a DATC implementation on a network of Windows NT workstations are used to determine the range of validity of the queuing model. The model is used to predict the latency of various platforms such as cluster of workstations, parallel computer, and reconfigurable computers.