To be updated ...

To be updated ...

Based on a distributed parallel framework, we propose an algorithm to evenly partition a forwarding table into lookup-independent sub-tables for subsequent parallel processing. The result of the algorithm is distinctly better than related works in terms of the balance of sub-table sizes and the amount of redundancy introduced during the partitioning. In addition, we also propose a cached-based approach to adaptively balance the real-time lookup traffic distribution among the parallel lookup engines.

We design a high-performance IPv6-compatible Packet Classification engine which can match 4*OC-768 line rate with relatively low hardware cost via efficient distributed parallel architecture. By incorporating a novel load-balancing scheme, the system performance only slides down slightly even under bursty traffic input pattern. We also propose search key encoding mechanisms to address the range matching problem in Packet Classification and to save memory cost.

We provide a fast pattern matching scheme which can be used in a signature-based Network Intrusion Detection System. First we propose to process multiple input characters within each memory clock by employing multiple input windows; then by exploiting the correlation between the input windows we enhance the parallel efficiency. We also build a CAM-based state machine to address the long or correlated signature matching problem.