An Effective-Traffic Based Dual-Threshold Queue Control Scheme in Shared Memory Switch

This paper investigates the queue control or buffer management problem in shared memory switches. Traditional methods such as dynamic threshold and the newly proposed such as decay function threshold take decision largely based on the current total queue length. The dual threshold method proposed by this paper considers the traffic scenario as the main factor to decide a per-port threshold. Each port maintains a common traffic intensity, which is accounted for by the traffic rate and the allotted buffer. The traffic intensity takes shape from the classical effective bandwidth theory, yet the formula devised by the theory is too complex to implement. This paper reduces the formula to a simple form that is eligible in the runtime environment with little loss of accuracy. Keeping the common traffic intensity by dynamically adjust the per-port threshold can maintain the system equilibrium and reach a global optimality. Simulation shows the proposed scheme outperforms the tradition dynamic thresholds significantly.