1. Pattern-Based DFA for Memory- Efficient and Scalable Multiple Regular Expression Matching Author: Junchen Jiang, Yang Xu, Tian Pan, Yi Tang, Bin Liu Publisher:IEEE International Conference on Communications (ICC) 2010 Presenter: Sih-An Pan Date: 2014/6/4 1 Department of Computer Science and Information Engineering National Cheng Kung University, Taiwan R.O.C.

2. Introduction Combining many regex patterns into a single DFA causes a serious state explosion. A practical solution of resolving the state explosion is to divide the set of regex patterns into several groups, and generate an individual DFA for each group. Computer & Internet Architecture Lab 2

3. Introduction A P-DFA records the information of every regex pattern in each of its state, and supports the operations like pattern insertion and deletion. We save such information in a data structure called Pattern-Based Structure (PBS). Computer & Internet Architecture Lab 3

4. 4

5. 5

6. Pattern-Based Structure (PBS) Computer & Internet Architecture Lab 6

7. Overview of Our Scheme Computer & Internet Architecture Lab 7

8. Overview of Our Scheme We define an operation called Reduce which takes two P-DFAs and one threshold L as input. By using the PBS to predict the size of the resting set, the operation always moves a pattern from one P-DFA to another to keep the size of the former as close as possible to L. Computer & Internet Architecture Lab 8

9. Design of Grouping Algorithms For general purpose processor architecture minimizes the number of groups while keeping the total number of states limited. After generating the complete P-DFA, Group1 keeps calling Reduce to split the current largest group in size into two small groups until the sum of all groups’ size is smaller than the given limit Computer & Internet Architecture Lab 9

10. Design of Grouping Algorithms For multi-parallel processor architecture minimizes the number of groups while keeping the number of states in each group limited. For one group, when the size of its P-DFA is larger than the limit, we extract a pattern from the group by calling Reduce, so that the size of P-DFA is as close to the limit as possible. Computer & Internet Architecture Lab 10

11. Efficiency of Pattern Grouping Algorithms Computer & Internet Architecture Lab 11

12. Efficiency of Pattern Grouping Algorithms Computer & Internet Architecture Lab 12