A Virtual Queue Approach to Loss Estimation

In network quality of service provisioning, small loss probabilities are often required. Such loss probabilities are so small that they may hardly be measured on line directly. In this paper, an indirect virtual queue (VQ) approach is proposed for loss estimation. The idea of the proposed approach is to use a scaling property between the real queue and a set of virtual queues to estimate the real loss probability from the loss probabilities in the virtual queues which can be feasibly measured. The scaling property is investigated on the basis of the maximum variance asymptotic (MVA) theory. The proposed approach like other VQ based approach has low implementation complexity. In addition, the analysis and results indicate that the proposed scheme retains the high accuracy of MVA while avoiding the high computational complexity in a direct application of the original MVA based estimation approach.