An Efficient TCAM Update Scheme for Packet Classification

Ternary Content Address Memory (TCAM) becomes a popular hardware device for storing the packet classifiers due to the advantages of high and deterministic lookup performance. However, managing the filter set in TCAM is quite complicated when filters need to be updated. To ensure the correctness of search results, it is required to obtain the right position in TCAM for storing the new filter. In addition, some filters need to be moved to other positions for maintaining the correct filter overlapping relationship based on their priorities. Normally, an auxiliary data structure is used to compute how to insert or delete a filter into/from TCAM. However, this auxiliary data structure is usually complicated and large. Instead of maintaining an auxiliary data structure, in this paper, we propose an efficient TCAM update scheme that simply uses a very small portion of TCAM search cycles to compute how to move the filters related to the filter to be inserted or deleted. Therefore, a large amount of memory for storing the auxiliary data structure along with the local CPU for updates can be avoided. In addition, our simulation results show that the proposed update scheme needs less number of TCAM movements than the existing CoPTUA update scheme [2].