Svante Ekelin

Real-Time Measurement of End-to-End Available Bandwidth using Kalman Filtering

This paper presents a new method, BART (bandwidth available in real-time), for estimating the end-to-end available bandwidth over a network path. It estimates bandwidth quasi-continuously, in real-time. The method has also been implemented as a tool. It relies on self-induced congestion, and repeatedly samples the available bandwidth of the network path with sequences of probe packet pairs, sent at randomized rates. BART requires little computation in each iteration, is lightweight with respect to memory requirements, and adds only a small amount of probe traffic. The BART method uses Kalman filtering, which enables real-time estimation (a.k.a. tracking). It maintains a current estimate, which is incrementally improved with each new measurement of the inter-packet time separations in a sequence of probe packet pairs. The measurement model has a strong non-linearity, and would not at first sight be considered suitable for Kalman filtering, but we show how this non-linearity can be handled. BART may be tuned according to the specific needs of the measurement application, such as agility vs. stability of the estimate. We have tested an implementation of BART in a physical test network with carefully controlled cross traffic, with good accuracy and agreement. Test measurements have also been performed over the Internet. We compare the performance of BART with that of pathChirp, a state-of-the-art tool for measuring end-to-end available bandwidth in real-time

Tuning the Temporal Characteristics of a Kalman-Filter Method for End-to-End Bandwidth Estimation

In this paper we present a way of tuning the temporal characteristics of a new available-bandwidth estimation method, BART. The estimation engine in this method is Kalman-filter based. A current estimate of the available bandwidth is maintained, and for each new sequence of probe packet pairs an updated estimate is produced. The main input parameters needed by the Kalman filter are the variance of the measurement noise and the covariance of the process noise. The former is measured by the method, whereas the latter is not in general attainable by analytical or empirical investigation. Instead, it is reasonable to treat this as a tunable parameter. We discuss how the temporal characteristics of the tracking of end-to-end available bandwidth may be tuned.

Real-time available-bandwidth estimation using filtering and change detection

This paper presents a filter-based method BART (Bandwidth Available in Real-Time) for real-time estimation of end-to-end available bandwidth in packet-switched communication networks. BART relies on self-induced congestion, and repeatedly samples the available bandwidth of the network path with sequences of probe-packet pairs. The method is light-weight with respect to computation and memory requirements, and performs well when only a small amount of probe traffic is injected. BART uses Kalman filtering, which enables real-time estimation. It maintains a current estimate, which is incrementally improved with each new measurement of the inter-packet time separation in a sequence of probe-packet pairs. It is possible to tune BART according to specific needs. The estimation performance can be significantly enhanced by employing a change-detection technique. An implementation of BART has been evaluated in a physical test network with carefully controlled cross traffic. In addition, experiments have been performed over the Internet as well as over a mobile broadband connection.

Enhanced Network-State Estimation using Change Detection

This paper presents the concept of change detection for filter-based network-state estimation. This could be useful in various contexts; two examples are network management and adaptive applications. In particular, it is shown that the performance of available-bandwidth estimation can be significantly enhanced by employing a change-detection technique in conjunction with a filter-based estimator. By using filter-based approaches, it is possible to track the state of communication systems, and to estimate network properties in real-time. A virtue of filter-based methods is the ability to enhance the estimation performance by combining them with change detection. This makes it feasible to overcome the tradeoffs regarding speed of adaptation to changes versus stable estimation. We discuss filtering and change detection in general, and illustrate the power of this combination with the filter-based available-bandwidth estimator BART enhanced by the light-weight change-detection test CUSUM

A Performance Study of Bandwidth Measurement Tools over Mobile Connections

The knowledge of the present available bandwidth on a network path is essential in numerous contexts, such as network management and streaming applications. A network path nowadays often contains at least one wireless link. This is obviously true for mobile users having a wireless connection to the Internet through a laptop or mobile terminal. The existing tools for measuring end-to-end available bandwidth are developed and optimized for paths with only guided media links. Since the characteristics for wired links and radio links differ in many aspects, such as fluctuations in capacity and stability, the network tools need to be evaluated also for network paths containing wireless links. In this investigation we have performed experiments over a high-speed downlink UMTS channel. This makes the present paper unique in the sense that it evaluates and analyzes the applicability of available-bandwidth measurement tools over a radio interface in a wide-area mobile communication network. For the experiments, a commercial mobile network has been used. The measurements show that it is feasible to achieve reliable estimates under certain circumstances. However, some cases pose challenges which motivate further studies.

Real-time bandwidth measurements over mobile connections

The first study that investigates the characteristics of received probe packets and the reliability of bandwidth estimates when actively measuring the available bandwidth over radio interfaces in mobile communication networks is presented. Knowledge of available bandwidth is very useful in various contexts, e.g. in network management and adaptive streaming applications. Bandwidth measuring tools have so far primarily been designed for and evaluated in wired networks. However, such tools should also be examined in wireless networks since the use of, e.g., mobile broadband is rapidly increasing. The properties of wired and wireless links differ substantially, which affect the performance of the tools. We have made active-probing experiments over a high-speed downlink shared channel, which is used for High-Speed Downlink Packet Access (HSDPA) in the mobile communication technology UMTS, and over a forward traffic channel in CDMA2000 1xEV-DO. Both experiments were performed over commercial networks. They show that one cannot always expect uniform per-packet processing over the radio channel in mobile networks, which is expected by many probing tools. This reduces the reliability of the available-bandwidth estimates, however we suggest how this can be handled. Finally, the mobile-network measurements are compared to experiments performed in an IEEE 802.11 wireless LAN, where the radio channel does not create the same packet-processing behaviour. We also discuss the possibility of using the probe traffic for the purpose of identifying the communication technology at the bottleneck of the network path, assumed this is a wireless broadband link, by mapping specifications of standardised communication technologies to observed probe-traffic characteristics. Copyright

Change detection and estimation for network-measurement applications

This paper presents a method for improving filter-based network-state estimation by adding detection and estimation of sudden changes in the system state. This can be of benefit in various contexts, e.g. in network management and adaptive streaming applications. In particular, it is shown that the performance of available-bandwidth estimation can be significantly enhanced by employing change detection in conjunction with a filter-based estimator. The use of filtering makes it feasible to track the communication network state and to estimate selected properties in real-time. In addition, filter-based methods may be combined with change detection in order to overcome the trade-off regarding stable estimation versus speed of adaptation to change. We discuss filtering and change detection in general, and present the novel approach of combining the filter-based available-bandwidth estimator BART with the Generalized Likelihood Ratio (GLR) change-detection test, which estimates both the time and magnitude of changes.

A scheme for measuring subpath available bandwidth

This paper presents a novel probing scheme which can be used for estimating the available bandwidth of subpaths, without the requirement of control over both endpoints of a network path. Instead of a probe-packet receiver, this scheme uses the ICMP capability of routers. An estimate of the available bandwidth from the endpoint to a router is obtained in much the same way as for state-of-the-art end-to-end probing methods. Taking into account ICMP packet generation limitations and delay, the estimate should be interpreted as a lower bound of the actual available bandwidth.