大学サイト   学部サイト   学科サイト   EnglishPages
 
旧尾家・塚本研究室です.尾家祐二 現・九州工業大学学長の情報は こちら.
   
研究業績


Framework for Efficient Estimation of QoS Parameters Associated with Paths through Multiple Routers
著者
  Nobuo Ryoki, Masato Tsuru, Yuji Oie
 
雑誌名/会議名
  Proceedings of International Conference on Networking ICN’2004, Vol II, pp.725-730, March 3, Gosier, Guadeloupe, French Caribbean 2004 年 3 月
 
アブストラクト
  End-to-end communication traverses multiple routers on a path, and thus QoS parameters associated with the path, as opposed to those seen independently at each router, are of major concern to Internet Service Providers (ISP) as well as end users. However, existing, common methods of estimating these QoS parameters (e.g., queuing delay and available bandwidth statistics) on individual paths among edge-routers in an ISP have certain drawbacks. Active measurement introduces wasteful traffic by sending probe packets along each path. In addition, it requires some calibration due to the difference in characteristics between user traffic and probe traffic. Passive measurement can also be employed to estimate edge-to-edge QoS parameters. One approach is independently measuring some states locally seen at each router and concatenating them along the path, while it may inaccurately estimate some statistics (rather than the mean) due to correlations among the local states at adjacent routers. The alternative is cooperatively sampling the same sets of packets at some routers to obtain QoS parameters seen by the samples. Sampling the same sets of packets at multiple routers is, however, costly so that a limited number of samples must be acceptable. This can further result in an inaccurate estimate. In this work, therefore, a general framework for efficiently, accurately and continually estimating edge-to-edge QoS parameters from two different kinds of data is proposed. Namely, local statistics measured independently at each router and a small number of samples over the path obtained cooperatively at each router are combined to compute a maximum likelihood estimator of the distribution of some joint-states associated with the path. Simulations are performed, which indicate that the proposed scheme can accurately estimate the corresponding QoS parameter compared with the existing methods, i.e., a simple concatenation based solely on the local statistics and a simple estimation based solely on the samples.
 
キーワード
  Measurement
 
記述言語
  English
 
 

web_admin [ at ] infonet.cse.kyutech.ac.jp /  Copyright (c) 2016 Oie & Tsukamoto Laboratory, All Right Reserved.