Andreas Johnsson Research Area Packet Systems Ericsson Research Mats Björkman The Department of Computer Science and Electronics Mälardalen University
Abstract— BART is a state-of-the-art active end-to-end bandwidth measurement method that estimates not only the available bandwidth but also the link capacity of the bottleneck link. It uses a Kalman ﬁlter to give estimates in real time during a measurement session. In this paper, we have studied the impact of 802.11 networks on the bandwidth estimates produced by BART. The Kalman ﬁlter used by BART is tunable, and one of the contributions of this paper is to show how the Kalman ﬁlter should be adjusted to improve real-time tracking and estimation accuracy when the bottleneck is an 802.11 link. Further, the paper contributes by discussing how to interpret the estimates produced by BART and similar bandwidth estimation tools relying on self-induced congestion when used in wireless scenarios. An analysis show that the BART estimates produced are correct - but corresponds to a fair share of the wireless link rather than to the unused capacity. However, the estimates do indicate how much bandwidth an application or device in the wireless network can expect when sending and/or receiving network trafﬁc.
I. I NTRODUCTION Since wireless networks often are used for nomadic access to the Internet from a laptop, it is important to study how methods for actively measuring available bandwidth behave when wireless links are part of the end-to-end path. If endto-end available bandwidth measurement methods are to be widely accepted as performance measurement tools, it is important to ensure that they report correct values in wireless networks as well. End-to-end available bandwidth methods relying on selfinduced congestion that exist today are for example ABget , Pathchirp , Pathload , Spruce  and TOPP . The principle is to inject so called probe packets, with some inter-packet separation, that traverse the network path to be measured. When a probe packet arrives at the receiver it is time stamped. The inter-packet separation has increased if the probe packets caused congestion on the path. By deploying different analysis methods to the sent and received time stamps the link capacity (a constant property of the bottleneck link) and/or the available bandwidth (the unused portion of the link capacity) can be estimated. The methods mentioned above differ in several ways, such as how the probe packets are
sent and how the analysis and estimation algorithms function. An overview of available bandwidth measurement methods, tools and theory can be found in . More recent theoretical ﬁndings are reported in for example  . BART  is a successor of the TOPP method and is used for actively estimating the end-to-end available bandwidth and link capacity on the IP layer. Assuming a ﬂuid network model and FIFO queues for the selected trafﬁc class, TOPP uses linear regression while BART deploys a Kalman ﬁlter to estimate the available bandwidth and link capacity. The advantage of using BART is that estimates are obtained in real time. To obtain fast-tracking and accurate bandwidth estimates using BART the Kalman ﬁlter can be tuned, this has been studied in several papers, for example in . In this paper tuning of the Kalman ﬁlter used by BART is examined in detail in scenarios where the bottleneck is an 802.11 link. For this study to be possible, a crucial observation is made in the paper: the estimated link capacity and available bandwidth are equal when using BART in scenarios where the bottleneck is an 802.11 link. The second contribution of this paper is a discussion of how to interpret the estimates produced by BART and similar tools when the bottleneck is an 802.11 link. An analysis show that the estimates produced are correct - but corresponds to a fair share of the wireless medium rather than to what was...