Split: Optimizing Space, Power, and Throughput for TCAM-Based Classification

Abstract

Using Ternary Content Addressable Memories (TCAMs) to perform high-speed packet classification has become the de facto standard in industry because TCAMs facilitate constant time classification by comparing packet fields against ternary encoded rules in parallel. Despite their high speed, TCAMs have limitations of small capacity, large power consumption, and relatively slow access times.One reason TCAM-based packet classifiers are so large is the multiplicative effect inherent in representing d-dimensional classifiers in TCAMs. To address the multiplicative effect, we propose the TCAM SPliT architecture, where a d-dimensional classifier is split into k ≥ 2 low dimensional classifiers, each of which is stored on its own small TCAM. A d-dimensional lookup is split into k low dimensional, pipelined lookups with one lookup on each chip. Our experimental results with real-life classifiers show that TCAM SPliT reduces classifier size by 84% using only two small TCAM chips; this increases to 93% if we use five small TCAM chips.

Scalable forwarding tables for supporting flexible policies in enterprise networks

“…We compare FISE and ACL-like in each step. We also compare with SPliT [17] structure, which first lookups in one dimension, outputs a sub-table, and merges sub-tables if they are the same. Because SPliT undergoes totally different steps, we only compare FISE and SPliT in the final step.…”

“…Most enterprise networks uses ACLlike structure, which is 'fat' in TCAM and 'thin' in SRAM [16]. However, TCAM-based solutions are limited by its capacity [17]. What is worse, TCAM is highly customized, there are limited techniques we can use to compress it.…”

PC-TRIO: A Power Efficient TCAM Architecture for Packet Classifiers

“…Liu et al [13], approach the range expansion problem by splitting a d-dimensional classifier into multiple lower dimensional classifiers each of which can be stored in its own small TCAM, leading to effective reduction of the range expansion problem. The algorithm effectively contains the range expansion problem, reduces power, improves the lookup time and supports batch updates.…”

“…In this 13. Data encoding in a wide SRAM word section we describe how these rules are stored in the LTCAM1 subsystem which comprises the LTCAM1 and LSRAM1, and the corresponding index structure ILTCAM1 and ILSRAM1.…”

A New Lookup Model for Multiple Flow Tables of Open Flow with Implementation and Optimization Considerations

“…The authors in [11] proposed the TCAM-based SPliT architecture where a d-dimensional classifier is split into k ( 2  k ) low dimensional classifiers, each of which is stored in its own small-size TCAM. A d-dimensional lookup is then equivalent to the k low dimensional and pipelined lookups with one lookup on each chip.…”