+ All Categories
Home > Documents > Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li...

Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li...

Date post: 14-Dec-2015
Category:
Upload: calista-brill
View: 216 times
Download: 2 times
Share this document with a friend
Popular Tags:
14
Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li ,Xianfeng Li Publisher :Engineering Lab on Intelligent Perception for Internet of Things (ELIP) Presenter : Kai-Hsun Li Date : 2014/12/24 Department of Computer Science and Information Engineering National Cheng Kung University, Taiwan R.O.C.
Transcript
Page 1: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Scalable Packet Classification Using Hybrid andDynamic Cuttings

  Authors : Wenjun Li ,Xianfeng Li  Publisher :Engineering Lab on Intelligent Perception for Internet of Things (ELIP) Presenter : Kai-Hsun Li  Date : 2014/12/24

Department of Computer Science and Information Engineering National Cheng Kung University, Taiwan R.O.C.

Page 2: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Introduction

Decision-tree based schemes are the most well-known solutions for packet classification.

HiCuts and HyperCuts, suffer from memory explosion problem.

EffiCuts , suffers from excessive memory accesses.

In this paper, propose HD-Cuts use hybrid and dynamic cuttings to improve memory accesses and performance simultaneously

National Cheng Kung University CSIE Computer & Internet Architecture Lab

2

Page 3: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Observation

National Cheng Kung University CSIE Computer & Internet Architecture Lab

3

Page 4: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(1/6)- Rule Set Partitioning(1/2)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

4

partition the rules into the following four subsets to separate large rules.

1) Subset_SA: rules with wildcard in SA.

2) Subset_DA: rules with wildcard in DA.

3) Subset_Small: rules with no wildcard in SA and DA.

4) Subset_Large: rules with wildcard in both SA and DA.

Page 5: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(2/6)- Rule Set Partitioning(2/2)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

5

thresh_large = 0.5

Subset_SA = {R1, R3, R5, R7, R9, R10, R13}

Page 6: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(2/6)- Rule Set Partitioning(2/2)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

6

thresh_large = 0.5

Subset_SA = {R1, R3, R5, R7, R9, R10, R13}

Subset_DA ={R2, R4, R6, R8}

Page 7: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(2/6)- Rule Set Partitioning(2/2)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

7

thresh_large = 0.5

Subset_SA = {R1, R3, R5, R7, R9, R10, R13}

Subset_DA ={R2, R4, R6, R8}

Subset_Small = {R11, R12}

Page 8: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(2/6)- Rule Set Partitioning(2/2)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

8

thresh_large = 0.5

Subset_SA = {R1, R3, R5, R7, R9, R10, R13}

Subset_DA ={R2, R4, R6, R8}

Subset_Small = {R11, R12}

Subset_Large = {R14}

Page 9: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(3/6)-Hybrid and Dynamic Cuttings (1/4)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

9

Subset_SA => step1. conduct HiCuts by considering SA、 DA. step2. if next cut is not “DA “ or rules in current node rules > binth

then extend to five field with HyperCut.

Subset_DA => step1. conduct HiCuts by considering SA、 DA. step2. if next cut is not “SA “

or rules in current node > binth then extend to five field with hypercut.

Subset_Small => step1.conduct HyperCuts by considering SA、 DA.step2.if rule replications become serious

or rules in current node rules > binth then extend to five field with HyperCut.

Subset_Large => HyperCut considering all five fields simultaneously.

Page 10: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(4/6)-Hybrid and Dynamic Cuttings (2/4)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

10

Subset_SA = {R1, R3, R5, R7, R9, R10, R13}

Page 11: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(5/6)-Hybrid and Dynamic Cuttings (3/4)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

11

Root(Hicuts)(DA_cuts:4)

Leaf 1R1

Leaf 2R3

Node 1R5,R7,R9,R13

Leaf 3R13

rules in current node rules > binth

=> extend to five field with hypercut

Page 12: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Proposed Scheme(6/6)-Hybrid and Dynamic Cuttings (4/4)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

12

Root(Hicuts)(DA_cuts:4)

Leaf 1R1

Leaf 2R3

Node 1(Hypercuts)(SA_cuts:1)(DA_cuts:1)(SP_cuts:2)(DP_cuts:2)(Port_cuts:1)

Leaf 3R13

Leaf 4R5

Leaf 5R7

Leaf 6R9

Leaf 7R10

Page 13: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Experiment Results(1/2)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

13

Total Memory for HiCuts, HyperCuts, EffiCuts and HD-Cuts(binth=8, thresh_large=0.05)

Page 14: Scalable Packet Classification Using Hybrid and Dynamic Cuttings Authors : Wenjun Li,Xianfeng Li Publisher : Engineering Lab on Intelligent Perception.

Experiment Results(2/2)

National Cheng Kung University CSIE Computer & Internet Architecture Lab

14

Memory Access for HiCuts, HyperCuts, EffiCuts and HD-Cuts(binth=8, thresh_large=0.05)


Recommended