Open-loop overload control algorithm for signalling traffic

Abstract

In this paper a novel algorithm is proposed to solve the problem of overload control in telephony signalling services. The new algorithm can provide better performance than the traditional ones and is developed for a control architecture containing a token bucket followed by a buffer. The behavior of the control loop is influenced by three parameters: (i) the token rate of a token-bucket restrictor denoted by r; (ii) the relative deadline of response times; and (iii) the zero point of the response times. Our concern is to optimize the control parameter r to maximize a utility function defined by two parameters: the relative deadline and the zero point of response times. The optimal control is sought by solving a constrained optimization problem in which we maximize the throughput under the constraint of tolerating a given response time. In this way, one can fulfill a pre-defined QoS criterion, while achieving optimal system performance. This optimization entails the evaluation of the queue length dynamics based on a 2D Markovian model. From this calculation the p.d.f. of the queue length and the system time is expressed and the QoS parameters are analytically calculated as a function of the control parameters. The optimal rate parameter can then be found by traditional optimization methods.