Brownian modeling of a rate control throttle

Congestion appearing in communication networks is typically caused by fluctuations in traffic input rates that cause offered traffic to a network resource to exceed the output capacity. Network protocols are saddled with the responsibility of controlling network traffic so as to maintain network performance at acceptable levels. One means of controlling congestion is to regulate the rate at which work is introduced into the network via a throttling mechanism. A rate-control throttle is employed to enforce a negotiated traffic contract, which specifies limits on the rate at which a traffic source can introduce traffic to a network. This mechanism throttles network traffic sources by delaying or blocking entry into the system when the source is deemed to not comply to its negotiated rate limits. Consequently, the statistical characteristics of the admitted traffic stream are altered in a complex manner.; To provide further insight into the behavior of the throttle's admitted process, we propose a Brownian model of the system. We formally define the rate control throttle model and show that it behaves like a well-known “two-sided regulator” model. We use stochastic calculus to compute various performance measures associated with the throttle. We are especially interested in measures associated with the admitted traffic stream. We compute the time average rate, time average variance, and effective bandwidths of the admitted traffic stream. Our analysis permits one to develop approximations for the queueing behavior at a downstream queue fed by throttled traffic sources.