Nowadays, using Deterministic Finite Automata (DFA) or Non-deterministic Finite Automata (NFA) to parse regular expressions is the most popular way for Deep Packet Inspection (DPI), and the research about DPI focuses ...Nowadays, using Deterministic Finite Automata (DFA) or Non-deterministic Finite Automata (NFA) to parse regular expressions is the most popular way for Deep Packet Inspection (DPI), and the research about DPI focuses on the improvement of DFA to reduce memory. However, most of the existing literature ignores a special kind of "overlap-matching expression", which causes states explosion and takes quite a large part in the DPI rules. To solve this problem, in this paper a new mechanism is proposed based on bitmap. We start with a simple regular expression to describe "overlap-matching expressions" and state the problem. Then, after calculating the terrible number of exploded states for this kind of expressions, the procedure of Bitmap-based Soft Parallel Mechanism (BSPM) is described. Based on BSPM, we discuss all the different types of "overlap-matching ex- pressions" and give optimization suggestions of them separately. Finally, experiment results prove that BSPM can give an excellent performance on solving the problem stated above, and the optimization suggestions are also effective for the memory reduction on all types of "overlap-matching expressions".展开更多
基金Supported by the National High Technology Development 863 Program of China (No. 2008AA01Z117)
文摘Nowadays, using Deterministic Finite Automata (DFA) or Non-deterministic Finite Automata (NFA) to parse regular expressions is the most popular way for Deep Packet Inspection (DPI), and the research about DPI focuses on the improvement of DFA to reduce memory. However, most of the existing literature ignores a special kind of "overlap-matching expression", which causes states explosion and takes quite a large part in the DPI rules. To solve this problem, in this paper a new mechanism is proposed based on bitmap. We start with a simple regular expression to describe "overlap-matching expressions" and state the problem. Then, after calculating the terrible number of exploded states for this kind of expressions, the procedure of Bitmap-based Soft Parallel Mechanism (BSPM) is described. Based on BSPM, we discuss all the different types of "overlap-matching ex- pressions" and give optimization suggestions of them separately. Finally, experiment results prove that BSPM can give an excellent performance on solving the problem stated above, and the optimization suggestions are also effective for the memory reduction on all types of "overlap-matching expressions".
