Measurement and Simulation Based Effective Bandwidth Estimation
Liu, Nelson X
Date: November 29 - December 03, 2004
Effective bandwidth represents the resource demand for network traffic to achieve its QoS goal. However, so far, the estimation of the effective bandwidth for modern Internet traffic has been difficult and inaccurate. We present a novel scheme using measurement and simulation techniques to estimate the effective bandwidth. The advantages of the scheme lie in its accuracy, its applicability to various traffic environments, and its efficiency. We describe the simulation procedure, the efficient search algorithm to lock the effective bandwidth quickly, the sample size determination algorithm, and the performance judgment policy in the simulation to guarantee the estimation accuracy. We also examine the issue of estimating the effective bandwidth of aggregate traffic from sub-aggregates, and provide an empirical formula to achieve very good accuracy.