Measurement-based virtual queue (VQ): to estimate the real-time bandwidth demand under loss constraint

Abstract

We study a traffic measurement issue for active queue management and dynamic bandwidth allocation at a single network node under the constraint of Cell Loss Probability (CLP) or buffer overflow probability. Using the concept of measurement-based Virtual Queue (VQ) and frequency-domain traffic filtering, we propose an online algorithm to estimate the real-time bandwidth demand under both short- and long-term cell loss constraint. The algorithm is adaptive and robust to the piece-wise stationary traffic dynamics. The VQ runs in parallel to the real queueing system and monitors the latter in a non-intrusive way. It captures a proper traffic sampling interval T c. Simulation and analysis show its critical role in achieving higher utilization of bandwidth and buffer resource without violating QoS requirement. Given an appropriate T c, we argue that network controls, such as the Call Admission Control (CAC) and dynamic bandwidth allocation, are facilitated with accurate information about traffic loading and QoS status at the smallest timescale.