Content Addressable Memory Decomposition for Cost Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ternary content-addressable-memory (TCAM) circuits are expensive and power inefficient due to the additional circuitry required for storing and searching ternary values, limiting their size and increasing costs compared to Random Access Memory (RAM).
Innovation Solution
Decomposing TCAM rules into smaller, lower-cost CAM circuits such as binary CAM (BCAM) or narrower CAM circuits configured for prefix match or exact match searches, where CAM rules are grouped and reordered to place wildcard bits in least significant positions, reducing the number of ternary bits required for storage and search operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If TCAM circuits are used to store and search ternary values with wildcard support, then search functionality and versatility are improved, but cost and power consumption increase significantly
Solution Approach 1:
The patent divides a large TCAM rule set into multiple subsets and further segments each subset into groups that can be processed by smaller CAM circuits. This segmentation allows the system to achieve TCAM-like functionality using multiple smaller, less expensive CAM circuits, thereby reducing overall cost while maintaining search versatility.
Solution Approach 2:
The patent introduces a rule reordering mechanism and search key formatting circuit as intermediary components. These intermediaries transform the input search keys and reorganize rule storage to enable wildcard matching functionality in standard CAM circuits, which normally lack this capability, thus achieving TCAM functionality without TCAM cost.
2Adaptability or versatility
If TCAM circuits are implemented with full ternary bit support, then wildcard matching capability is improved, but die area and memory requirements increase
Solution Approach 1:
The patent changes the dimension of the problem by transforming the storage structure from full N-bit ternary representation to a segmented approach where rules are reordered and stored in smaller CAM circuits. This dimensional change in data organization enables wildcard matching without requiring full TCAM die area.
Solution Approach 2:
The patent creates reformatted copies of the original rule sets in smaller CAM circuits. By copying and reorganizing rules into subsets with reordered bits, the system achieves wildcard matching capability using multiple smaller copies rather than a single large TCAM, reducing overall die area.
3Loss of information
If CAM rules are stored in full N-bit format, then complete information is preserved, but storage cost and circuit complexity increase
Solution Approach 1:
The patent performs preliminary reordering of rule bits during the storage configuration phase. By pre-reordering bits within groups to place wildcard bits in consistent positions, the system enables efficient wildcard matching in smaller CAM circuits without losing any rule information, thereby reducing storage cost while maintaining completeness.
4Device complexity
If smaller CAM circuits are used instead of full TCAM, then cost is reduced, but search capability for ternary values is limited
Solution Approach 1:
The patent makes standard CAM circuits universal by introducing a search key formatting circuit that transforms input keys and a rule reordering mechanism. This allows the smaller CAM circuits to perform multiple functions including exact matching, prefix matching, and wildcard matching, thereby achieving multi-functionality without increasing circuit cost.
Data Source
AI summary
Circuits and methods are disclosed for decomposition of a content addressable memory into a plurality of CAMs having a lower cost. In an example implementation, a set of CAM rules are grouped into a plurality of subsets. For each of the subsets, CAM rules in the subset are reformatted for storage in a respective CAM configured to store fewer ternary bits or configured for prefix match. Each reformatted subset of CAM rules are stored in the respective CAM. A search key formatting circuit is configured to reformat an input search key for each of the respective CAMs is used to store the reformatted subsets to produce a respective reformatted search key and input the respective reformatted search key to the respective CAM.


