« 2007年03月07日 | メイン | 2007年03月09日 »

2007年03月08日

Real-Time Traffic Classification Research

"Traffic Classification on the Fly," ACM SIGCOMM CCR Vol. 36 , Issue 2 (April 2006). Machine Learning による IP フロー分類の研究。フローの最初の 5 パケットだけを用いて分類を行う Simple K-Means クラスタリングアルゴリズムを使用。

JSSST コンピュータソフトウェア誌 Research

3 年連続。校正・編集は岩波書店枚数制限がきつい。。

"SPA 2006 の報告," 斉藤匡人, 石川宗寿, 川添瑞木, 栗田洋輔, 小林思無邪, 津田恵理子, 堀江倫大 日本ソフトウェア科学会コンピュータソフトウェア, Vol. 23, No. 4, pp. 84-93, 2006 年 10 月

NetMate Meter Research

NetMate Meter is a flexible and extensible tool for network measurement (packet capturing). It is the meter which listens on network interfaces, classifies packets and computes metrics.

Metric Modules - port usage, packet ID generation (crc32 and md5), RTT (ICMP echo), DNS latency, HTTP performance, TCP connection setup latency など。

NetMate (Network Measurement and Accounting System). UNIX 向けの C++ 実装(C++ classes, GPL)。特徴は、XML-based Ruleset, IPv6 Support, Multiple Packet Classification Algorithms, Packet Sampling Support など。

10 Papers on Network Models Research

"10 Papers on Network Models (Matthias Grossglauser)," ACM SIGCOMM CCR Vol. 36, Number 5 (October 2006). Models allow us to reason about scale before we are able or willing to build a large-scale artifact. Networking is an engineering discipline. Often, major new applications did not even emerge from academic projects, but were developed by creative young people outside the research community.

"On the Self-Similar Nature of Ethernet Traffic," [W. E. Leland 1994] (a baffling observation at the time).
"Network Tomography: Estimating Source-Destination Traffic Intensities from Link Data," [Y. Vardi 1996] (the first paper that formulates the estimation of a traffic matrix (the amount of traffic between source-destination pairs), PAM 研究の走り).
"Linear Network Coding," [S.-Y. R. Li 2003] (introduces the notion of network coding).

MLs for Flow Classification Research

"A Preliminary Performance Comparison of Five Machine Learning Algorithms for Practical IP Traffic Flow Classification," ACM SIGCOMM CCR Vol. 36, Number 5 (October 2006). port number & payload ベースのフロー識別ではない、payload-independent なパケット長や到着時間間隔分布など(features)をベースにしたフロー統計によるフロー識別方法の性能比較。The same set of features (この論文では 22 の flow features) をフロー群に適用し、フローごとに異なる feature values によってフローを識別。Instances (unknown flows) を classes (known flows) にマッピングする(基本的に MLs は supervised algorithms = 事前にマッピング情報を持っている)。

研究背景は、port-agile, using non-default ports, using network address port translation なアプリケーションの増加。比較した ML 群は、Bayesian Network, C4.5 Decision Tree, Naive Bayes, Naive Bayes Tree (実装がシンプルで、パラメータ調整が必要ないものを選択). ML 群の分類の正確さはほぼ同じ、build time (required to train a classifier) や分類スピードなどの計算性能が ML ごとに大きく異なる。評価結果は、C4.5 の分類スピードが最も早い。

A Day in the Life of the Internet Research

"A Day in the Life of the Internet: Proposed community-wide experiment (kc claffy)," ACM SIGCOMM CCR Vol. 36, Number 5 (October 2006).

The outsiders expressed the view that the network research community should not devote all or even the majority of its time to fixing current Internet problems. Instead, networking research should more aggressively seek to develop new ideas and approaches; measurement of the Internet, modeling of the Internet, and making disruptive prototypes. This report challenges the research community to develop the means to capture a day in the life of the Internet.

The Internet lacks the means to perform comprehensive measurement on activity in the network. The proposed project will involve building a cooperative community to support the simultaneous capture of a variety of measurements from and across many strategic links around the globe by supporting tools, analysis, visualization, and data catalog (DatCat, Internet Traffic Archive, CRAWDAD, MOME (Measurement Tools), Datapository, PREDICT).

Domenico Ferrari Research

"Domenico Ferrari Wins 2006 ACM SIGCOMM Award (Ellen Zegura)," ACM SIGCOMM CCR Vol. 36, Number 5 (October 2006). 現在は、Università Cattolica, Piacenza, Italy に。

In recognition of fundamental contributions to Internet QoS architecture (a pioneer in the area of real-time communications) and wide-area network research testbeds, his leadership in managing the BSD Unix project (he led the computer research systems group at Berkeley which developed BSD Unix), his leadership in founding and establishing ICSI as a premier international networking research institutions.

2006 SIGCOMM Test of Time Award Research

The 2006 ACM SIGCOMM Test of Time Award has been given to Vern Paxson for his paper, “End-to-End Routing Behavior in the Internet,” published in the 1996 proceedings of the ACM SIGCOMM Conference.

It marks a moment of change in network measurement (a new kind of network measurement). Paxson showed that, using proper statistical techniques, one could gather considerable information about the behavior of the network core using measurement stations solely at the edge of the network (40,000 end-to-end route measurements conducted using repeated “traceroutes” between 37 Internet sites).

The experimental methodology is clearly spelled out, such that the experiment is repeatable by someone else.

In 1969, when the ARPANET was being built, Len Kleinrock at UCLA was commissioned to put together a measurement center to analyze the performance of the network. Network Measurement の歴史

DOME TOUR 2005 "I ♥ U" Music

LIVE DVD MR.CHILDREN DOME TOUR 2005 "I ♥ U" ~FINAL IN TOKYO DOME~。ライブ用アレンジ感動的な [Monster] [Worlds end] [未来] のフル視聴ストリーミングを配信中。素晴らしいクオリティ

MR.CHILDREN DOME TOUR 2005
トイズファクトリー (2006/05/10)