Abstract
With the exponential growth of the Internet and various services provided, several requirements such as the next generation Internet Protocol (Internet Protocol Version 6) and the technique of classifications are needed. However, to achieve the requirements described above, there are two issues. First, filters in multi-field classifiers have more complex overlapping situations than in single-field. The thesis proposes an algorithm to detect the overlapped filters in multi-field classifiers and judge the overlapping situations when adding a filter. Second, the processing time of classifications is according to the total length of all fields and the number of filters in classifiers. Higher layer classifications take more time to classify packet flows.A Ternary content-addressable memory (TCAM) is a hardware solution for classifying packet flows. TCAMs perform high-speed parallel search operations to promote the performance of classifiers. But the hardware architecture of TCAMs leads to two problems. First, filters inserted to TCAMs should be in the form of prefixes. To solve the problem, this thesis designs a method to translate from a range into prefixes. Second, locations in TCAMs and priorities of filters have relations. The relations cause a lot of updates to keep entries in decreasing order of priority in a classifier and bring about low performances. This thesis proposes a flexible and optimal method to update the TCAMs based classifiers, and gives simulation results to prove the practicability.