Power-Efficient Rule Table Partitioning Algorithms Based on TCAM
碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 95 === Packet classification is a key technology that would affect the total system performance of modern routers. Many researches have widely studied this problem by using TCAM devices to speedup the search process. However, the usage of TCAM suffers from the issue...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2007
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/14422851024757767567 |
| Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 95 === Packet classification is a key technology that would affect the total system performance of modern routers. Many researches have widely studied this problem by using TCAM devices to speedup the search process. However, the usage of TCAM suffers from the issue of high power consumption such that the power budget gets large. Also since TCAM devices can not efficiently represent the arbitrary ranges such that each rule in rule table may occupy many number of TCAM entries. Due to the property that partial TCAM entries can be enabled at a time, Extended TCAMs [16] use the recursive cut to partition the multidimensional space to save the average power consumption. It also employs particular hardware designs to solve the inefficient representation of arbitrary ranges in TCAMs which requires additional power budget. In this paper, we design efficient partitioning algorithms to reduce the average power consumption while using TCAMs and improve the efficiency of the representation for arbitrary ranges by using the existing encoding technique. Besides, we develop parallel architectures which can truly reduce the total TCAM size. We use ClassBench [19] to generate rule tables and trace files with different properties to test our proposed algorithms and architectures. The results show that our methods can reduce the power consumption by a significant factor compared with the conventional TCAM.
|
|---|