Wenwei Li

Design and Implementation of a Network Behavior Analysis-Oriented IP Network Measurement System

Analyzing the characteristic of network behavior provides scientific basis for designing, building, and managing the next generation Internet, and is especially important for monitoring network behavior. This paper establishes a system of metrics that evaluates the behavior of IP networks with respect to the need of analyzing network behavior, introduces the design and implementation of network monitoring system that focuses on the analysis of the characteristics of network behavior, analyzes crucial problems on system design, builds an experiment environment and runs tests on it. The results show that our system satisfies all requirements imposed by real time monitoring network behavior, therefore is able to help the decision making in operating and managing networks.

An Adaptive Sampling Methodology for Internet Traffic Data Measurement

Current popular sampling algorithms can’t provide accurate approximation of true character under the circumstance of network traffic bursting. In the paper, we develop an adaptive network traffic sampling methodology(MANFC-AT) based on aggregated time-serial evolution trend.The MANFC-AT algorithm adaptively adjusts the time interval between consecutive samples according to the evolution trend of network traffic which is estimated from the obtained samples. Then it is able to approximate the measured traffic more precisely. Compared with current popular sampling algorithms, MANFC-AT algorithm can provide more precise estimation of network traffic, only requires fewer samples,contains more information about network traffic bursting.

Performance evaluation of end-to-end path capacity measurement tools in a controlled environment

In this paper we present results of a series of end-to-end path capacity measurement experiments conducted on a fully controllable IP network built in the lab. We test and evaluate several typical path capacity measurement tools: pathchar, clink, and pathrate. The cross traffic is generated in a reproducible manner, and has similar spatial and temporal characteristics of real Internet traffics. We evaluate the measurement error and other operational characteristics of the tools and analyze factors impacting their performance. Such work is useful for tools selection and further improvement of capacity measurement tools.

TCP and ICMP in network measurement: an experimental evaluation

Both TCP and ICMP are applied in network measurement, while investigating differences between the measured results of them is important but has been less addressed. To compare the differences between TCP and ICMP when they are used in measuring host connectivity, RTT, and packet loss rate, we designed two groups of comparison programs, after careful evaluating of the program parameters, we executed a lot of experiments on the Internet. The experimental results shows, there are significant differences between the host connectivity measured using TCP or ICMP; in general, the accuracy of TCP is 20%-30% higher than that of ICMP. The case of RTT and packet loss rate is complicated, which are related to path loads and destination host loads. While commonly, the RTT and packet loss rate measured using TCP or ICMP are very close. We also give some advices on protocol selection for conducting accurate network measurements.

Measurement and Evaluation Performance of Stream-Media Service Based on Active Measurement

Stream-media technology is a hotspot of research these days. The performance of stream-media service provided by the Internet based on TCP/IP protocol is always changing with the variation of network and is easily affected by the network parameters. During the developing of stream media, service performance Measurement is very important. This paper proposes parameter definitions to measure performance, from a user perspective, of stream-media, develops an appropriate measurement tool based on active measurement technology that focus on these agreed upon list of parameters, and builds a stream-media service performance test bed that allows us to systematically vary network characteristics. Then we analyze the service performance behavior under various streaming configuration and network conditions. While there has been research done in video measurement and characterization using custom tools, to the best of our knowledge, there have been no performance measurement studies where the researchers had control of the popular stream-media server, content parameters and network conditions between client and server.

A resource allocating neural network based approach for detecting end-to-end network performance anomaly

Automatic detection of end-to-end network performance anomalies is important to efficient network management and optimization. We present an end-to-end network performance anomalies detection method, based on characterizing of the dynamic statistical properties of RTT normality. The experiment on real Internet end-to-end path RTT data shows that, the proposed method is accurate in detecting performance anomalies, it can successfully detect about 96.25% anomalies in the experiment.

An active measurement approach for link faults monitoring in ISP networks

Link failures are common in IP networks. In this paper, we develop a technique for diagnosing link failures in Service Provider or Enterprise IP networks using active measurement.The technique attempts to minimize both the monitoring infrastructure costs as well as the additional traffic due to probe messages. Both the station selection problem and the probe message assignment problem are NP-hard, and we propose approximation algorithms for these problems. We also apply our algorithms to existing ISP networks using topologies inferred by the Rocketfuel tool. The results can be used by ISP operators to trade off between accuracy of diagnosis and monitoring costs.