Regression-Based Available Bandwidth Measurements∗

In this paper we present a method for estimating the available bandwidth of a network path. It is an extension and enhancement of the bandwidth measurement method TOPP. TOPP actively probes a network path by sending probe packets in a predetermined time pattern. Our enhancement involves a formalized estimation algorithm based on constrained linear regression. Using the algorithm, bandwidth measurements can be fully automated requiring no assistance from the user. We show that our method is able to estimate bottlenecks that cannot be detected by packet train methods such as C-probe. In addition to inferring the available bandwidth, the method gives an estimate of the link bandwidth of the most congested link on the network path. The link bandwidth estimates are not limited to the rate at which we can inject probe packets into the network.