Predicting queue size to avoid message lose in service discovery protocols

The use of broadcast based, service discovery protocols puts a significant strain on the delivery reliability of messages, particularly when a large number of devices simultaneously issue discovery requests. This usually results in undelivered discovery requests either requiring the requests to be issued again or the services remaining undiscovered. One of the principle courses of such message loss is the queuing mechanism used by the intermediate routers and/or any control points required to manage devices. This paper presents a significant analysis of the relationships between the different network devices to try to understand the causes of such message loss. In particular it will examine the different causes and how they influence each other. We present a formula which shows the relationship between the different causes of undelivered messages. This formula has been applied to the Universal Plug and Play (UPnP) protocol using simulated client and service applications written in C++. This paper provides researchers and network designers a mechanism for analyzing the potential performance of a network. It allows the impact of the relationships between the different causes for the non-delivery of messages to be taken into consideration when designing a network and applying protocols to that network.