Regenerative approach for effective bandwidth simulation

The Effective Bandwidth (EB) estimation problem has received much attention nowadays. EB is the powerful metric using for admission control in communication networks and related to QoS requirements assigned with loss probability and waiting time delay. For given arrival process EB allows to determine the amount of resources which is necessary to provide a required QoS level. In particular, for buffered stationary queue with service constant rate C the EB problem is to find out the minimal service rate which will ensure an overfull probability less than given value Γ.

The large deviation theory gives us exponential relation be- tween the buffer size and overfull probability. The EB estimation in turn is reduced to estimation of Scaled Cumulant Generating Function (SCGF). In the simplest case the independence of arrival work {v_n, n ≥ 1} per time unit is supposed. But, if the arrival se- quence contains dependences the special approach for SCGF eval- uating is necessary.

This work presents the regenerative method for EB estima- tion and their advantages in comparison with the traditional batch means methods. Besides, refined regenerative approach allows to receive conservative (pessimistic) EB estimator, that is important from the practical point of view.

The work is supported by the Russian Foundation for the Basic Research, Grant 10-07-00017.