介紹了流分類算法的概念以及對(duì)流分類算法的要求；把目前存在的流分類算法分成三類：多維查找轉(zhuǎn)換為一維查找算法、相關(guān)區(qū)域查找算法、獨(dú)立區(qū)域查找算法，并對(duì)各類算法的性能進(jìn)行了討論；通過(guò)引入并行流分類算法說(shuō)明了流分類算法的研究重點(diǎn)是減小存儲(chǔ)空間和提高更新速度。
關(guān) 鍵 詞 流分類; 多維查找轉(zhuǎn)換為一維查找算法; 相關(guān)區(qū)域查找算法; 獨(dú)立區(qū)域查找算法; 并行流分類

隨著英特網(wǎng)的發(fā)展，流分類技術(shù)得到廣泛地應(yīng)用，如防火墻、基于策略的路由、流量控制、流量整形、計(jì)費(fèi)等等。與流分類有關(guān)的策略和規(guī)則的集合稱為流分類器或簡(jiǎn)稱分類器。分類器中的每一條規(guī)則根據(jù)數(shù)據(jù)包頭的某些域來(lái)定義數(shù)據(jù)包所屬的流。常用的幾個(gè)域有：目的IP地址、源IP地址、源端口、目的端口、協(xié)議類型等。

Abstract The conception of packet classification and request for it is introduced. The paper classifies the existed packet classification algorithms into three kinds: conversion into single-field search、dependent field search、independent field search, then compare their performance. The presentation of parallel packet classification indicates how to reduce the storage space and update time is the research importance.
Key words packet classification; conversion into single-field search; dependent field search; independent field search; parallel packet classification