End-to-end loss and delay are both fundamental metrics in network performance evaluation, and accurate measurements for these end-to-end metrics are one of the keys to keeping delay/loss-sensitive applications (e.g., audio/video conferencing, IP telephony, or telesurgery) comfortable on networks. In our previous work [1], we proposed a parallel flow monitoring method that can provide accurate active measurements of end-to-end delay. In this method, delay samples of a target flow increase by utilizing the observation results of other flows sharing the source/destination with the target flow. In this paper, to improve accuracy of loss measurements, we propose a loss measurement method by extending our delay measurement method. Additionally, we improve the loss measurement method so that it enables to fully utilize information of all flows including flows with different source and destination. We evaluate the proposed method through theoretical and simulation analyses. The evaluations show that the accuracy of the proposed method is bounded by theoretical upper/lower bounds, and it is confirmed that it reduces the error of loss rate estimations by 57.5% on average.

In this paper, a parallel flow monitoring technique that achieves accurate measurement of end-to-end delay of networks is proposed. In network monitoring tasks, network researchers and practitioners usually monitor multiple probe flows to measure delays on multiple paths in parallel. However, when they measure an end-to-end delay on a path, information of flows except for the flow along the path is not utilized in the conventional method. Generally, paths of flows share common parts in parallel monitoring. In the proposed method, information of flows on paths that share common parts, utilizes to measure delay on a path by partially converting the observation results of a flow to those of another flow. We perform simulations to confirm that the observation results of 72 parallel flows of active measurement are appropriately converted between each other. When the 99th-percentile of the end-to-end delay for each flow are measured, the accuracy of the proposed method is doubled compared with the conventional method.

For network researchers and practitioners, active measurement, in which probe packets are injected into a network, is a powerful tool to measure end-to-end delay. It is, however, suffers the intrusiveness problem, where the load of the probe traffic itself affects the network QoS. In this paper, we first demonstrate that there exists a fundamental accuracy bound of the conventional active measurement of delay. Second, to transcend that bound, we propose INTrusiveness-aware ESTimation (INTEST), an approach that compensates for the delays produced by probe packets in wired networks. Simulations of M/M/1 and MMPP/M/1 show that INTEST enables a more accurate estimation of end-to-end delay than conventional methods. Furthermore, we extend INTEST for multi-hop networks by using timestamps or multi-flow probes.

To efficiently monitor the link performance in an OpenFlow network with a single measurement box (referred to a “beacon”), this paper presents a measurement scheme that calculates a set of measurement paths from the beacon to cover all links in the network based on the controllable feature of individual measurement paths in the OpenFlow network and comprehensively estimates the performance of all the physical links from round-trip active measurements. The scheme has a novel feature that minimize the maximum number of exclusive flow-entries for active measurements on OpenFlow switches by utilizing common packet header values in the probing packets to aggregate multiple entries into a single entry to save the resources in OpenFlow switches and controller. We demonstrate the effectiveness and feasibility of our solution through simulations and emulation scenarios.

The Internet real-time applications are growing rapidly, and available bandwidth estimation is required. Available bandwidth estimation methods by end host have been studied e.g. Pathload and pathChirp. These methods parameterize probe packet volume and observe the delay variation to estimate available bandwidth. In these methods, the probe packets impose heavy overhead loads on the network. In this paper, we propose a new available bandwidth estimation method based on the frequency of minimum RTT of probe packets in multi hop links. This method estimates bandwidth utilization and available bandwidth of a bottleneck link without significantly increasing network overhead. Estimation accuracies are evaluated for available bandwidth by implementing the proposed method. The proposed method shows better performance than pathChirp or Pathload, requiring fewer probe packets and less estimation time simultaneously.

Active measurement is an end-to-end measurement technique that can estimate network performance. The active measurement techniques of PASTA-based probing and periodic-probing are widely used. However, for the active measurement of delay and loss, Baccelli et al. reported that there are many other probing policies that can achieve appropriate estimation if we can assume the non-intrusive context (the load of the probe packets is ignored in the non-intrusive context). While the best policy in terms of accuracy is periodic-probing with fixed interval, it suffers from the phase-lock phenomenon created by synchronization with network congestion. The important point in avoiding the phase-lock phenomenon is to shift the cycle of the probe packet injection by adding fluctuations. In this paper, we analyse the optimal magnitude of fluctuations corresponding to the given autocovariance function of the target process. Moreover, we introduce some evaluation examples to provide guidance on designing experiments to network researchers and practitioners. The examples yield insights on the relationships among measurement parameters, network parameters, and the optimal fluctuation magnitude.

This paper presents an IP performance management system having the triple frameworks of performance measurement, topology monitoring and data analysis. The system infers the causal location of the performance degradation with a network tomographic approach. Since the Internet is still highly prone to performance deterioration due to congestion, router failure, and so forth, not only detecting performance deterioration, but also monitoring topology and locating the performance-degraded segments in real-time is vital to ensure that Internet Service Providers can mitigate or prevent such performance deterioration. The system is implemented and evaluated through a real-world experiment and its considerable potential for practical network operations is demonstrated.

Since congestion is very likely to happen in the Internet, locating congested areas (path segments) along a congested path is vital to appropriate actions by Internet Service Providers to mitigate or prevent network performance degradation. We propose a practical method to locate congested segments by actively measuring one-way end-to-end packet losses on appropriate paths from multiple origins to multiple destinations, using a network tomographic approach. Then we conduct a long-term experiment measuring packet losses on multiple paths over the Japanese commercial Internet. The experimental results indicate that the proposed method is able to precisely locate congested segments. Some findings on congestion over the Japan Internet are also given based on the experiment.

This paper proposes ABdis, a new active measurement method for estimating the available bandwidth on a communication network path. Due to the recent explosion in multimedia applications, it is becoming increasingly important to manage network QoS, and tools that can measure network quality precisely are necessary to do this. Many conventional active measurement methods/tools, however, can measure/estimate only the average of the available bandwidth in a given period but cannot measure its distribution. If the distribution of the bandwidth over short intervals can be measured, the information would be useful for network management, proxy selection, and end-to-end admission control. We propose an end-to-end active measurement method called ABdis, which can estimate the distribution of the available bandwidth in a network path. ABdis employs multiple probes having different rates and a parameter-matching technique for estimating distribution. Furthermore, we present the results of simulations and verify ABdis's performance under various conditions by changing probe parameters, amount of cross traffic, and network models.

Since the TCP is the transport protocol for most Internet applications, evaluation of TCP throughput is important. In this paper, we establish a framework of evaluating TCP throughput by simple measurement. TCP throughput is generally measured by sending TCP traffic and monitoring its arrival or using data from captured packets, neither of which suits our proposal because of heavy loads and lack of scalability. While there has been much research into the analytical modeling of TCP behavior, this has not been concerned with the relationship between modeling and measurement. We thus propose a lightweight method for the evaluation of TCP throughput by associating measurement with TCP modeling. Our proposal is free from the defects of conventional methods, since measurement is performed to obtain the input parameters required to calculate TCP throughput. Numerical examples show the proposed framework's effectiveness.