HTTP based communication protocols are ubiquitous in modern distributed control/supervision systems, therefore it becomes important to understand the performance of the HTTP protocol working with this type of applications and to develop techniques to investigate the impact of the protocol on the overall performance of these systems. A distributed control/supervision system based on HTTP usually consists of multiple nodes sending control requests or measurements to one or more recipient nodes; these data are used to perform control functions, and are stored or presented to supervising personnel. Typically multiple, concurrent flows of data are carried over HTTP to a given recipient. The stochastic characteristics of these flows are often different from the bursty nature of the ordinary Web Browser – Web Server interactions. Furthermore, in order to efficiently carry the randomly arriving data, the number of overlapping (concurrent) HTTP request/response sessions utilized by a sending node is higher than in typical Web Browser -- Web Server communications.

We investigate how the characteristics of nodes participating in the distributed control system communications (number of nodes, service times, number of service stations, queue lengths) influence performance of the employed HTTP protocol. The main focus is on  end-to-end delays in transmissions, the most important factor in control. The presented approach can be used to select configuration parameters for the nodes of a realistic control/supervision system having in mind the best communication related performance of such system.

In this work we propose a queuing model of HTTP client/server interactions which we subsequently simplify to be represented by a numerically solvable Markov chain. We compare the results obtained with the simplified model with the output of a discrete event simulation based on the original full model.

The initial motivation for this research came from a study of UPnP based remote PTZ camera control mechanism used by a leading manufacturer of large scale video monitoring systems.