Presentation 2017-05-12
An RBT Decision Tree Construction for Sparse Rules
Takashi Harada, Ken Tanaka, Kenji Mikawa,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Packet classification problem is to determine the highest priority rule that matches with an incoming packet in a networking equipment among all the filtering rules implementing a packet classification policy. Since the number of rules $n$ in a rule list tends to increase, it is necessary to develop an algorithm to find the highest priority rule independent of $n$. Thus we proposed a data structure Run-Based Trie (RBT) and a classification algorithm using decision tree constructed on RBT[1],[2]. The decision tree classification method is independent of $n$. However, a space complexity of the decision tree is exponential order $O(n^{w})$ with $n$ rules of length $w$. Even for Rules with $w geq 18$ and sparse rules generated by ClassBench[3] the decision tree has not been constructed. In this paper, we propose a novel decision tree construction method that does not construct unnecessary nodes by referencing runs matched through the root node to a node $v$ when constructing a child node of $v$. The proposed method can construct the decision tree for sparse rules generated by ClassBench.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) packet classification / trie / decision tree / arbitrary bitmask
Paper # COMP2017-2
Date of Issue 2017-05-05 (COMP)

Conference Information
Committee COMP / IPSJ-AL
Conference Date 2017/5/12(2days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair Hiro Ito(Univ. of Electro-Comm.) / 堀山 貴史(埼玉大)
Vice Chair Yushi Uno(Osaka Pref. Univ.)
Secretary Yushi Uno(Seikei Univ.) / (Kyushu Inst. of Tech.)
Assistant

Paper Information
Registration To Technical Committee on Theoretical Foundations of Computing / Special Interest Group on Algorithms
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) An RBT Decision Tree Construction for Sparse Rules
Sub Title (in English)
Keyword(1) packet classification
Keyword(2) trie
Keyword(3) decision tree
Keyword(4) arbitrary bitmask
1st Author's Name Takashi Harada
1st Author's Affiliation Kanagawa University(Kanagawa Univ.)
2nd Author's Name Ken Tanaka
2nd Author's Affiliation Kanagawa University(Kanagawa Univ.)
3rd Author's Name Kenji Mikawa
3rd Author's Affiliation Niigata University(Niigata Univ.)
Date 2017-05-12
Paper # COMP2017-2
Volume (vol) vol.117
Number (no) COMP-28
Page pp.pp.9-15(COMP),
#Pages 7
Date of Issue 2017-05-05 (COMP)