TCAM LPM Compression Using Hash-Based Parallel Prefix Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Trie search techniques used in ternary content addressable memory (TCAM) for determining longest prefix matches in IP addresses are inefficient, leading to long and non-deterministic delay times, and occupy memory resources that cannot be used for alternative purposes.
Innovation Solution
A modulo math-based hash function is used to determine memory addresses for IP address keys, allowing simultaneous parallel comparison of non-X values to find the longest prefix match, eliminating the need for trie tree traversal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If trie search technique is used for LPM algorithm, then IP prefix matching can be performed, but delay time increases and becomes non-deterministic
Solution Approach 1:
The patent segments the IP address key into multiple fields (e.g., first field, second field, third field) and stores them in separate memory locations. This segmentation allows parallel access to different parts of the IP address, enabling simultaneous comparison operations that reduce search delay while maintaining deterministic timing.
Solution Approach 2:
The patent transitions from a sequential tree-based search approach to a parallel memory-based search approach. By organizing IP prefix data in a multi-dimensional memory structure with separate fields and locations, the system enables parallel comparison operations instead of sequential traversal, significantly reducing delay time while maintaining matching accuracy.
2Measurement precision
If trie search technique is used for LPM algorithm, then IP prefix matching can be performed, but device complexity increases
Solution Approach 1:
The patent divides the IP address into multiple segmentable fields (first field, second field, third field, etc.) that can be stored and processed independently in memory. This segmentation simplifies the memory structure by allowing straightforward field-based organization rather than complex tree structures, reducing device complexity while maintaining matching precision.
Solution Approach 2:
The patent reorganizes data storage from a hierarchical tree structure to a flat, field-based memory structure. By using separate memory locations for different IP address fields and enabling parallel access, the system reduces structural complexity while achieving the same LPM functionality through dimensional reorganization of data storage.
3Measurement precision
If trie search technique is used for LPM algorithm, then IP prefix matching can be performed, but memory resources cannot be used for alternative purposes
Solution Approach 1:
The patent segments IP address data into discrete, field-based units that can be independently stored and managed in memory. This segmentation allows the memory structure to be more flexible and adaptable, as each field can be independently accessed, modified, or repurposed for alternative uses while maintaining the core LPM functionality.
Solution Approach 2:
The patent transforms the memory organization from a rigid, dedicated tree structure to a flexible, field-based structure that supports parallel access and alternative uses. By organizing data in separable fields with independent memory locations, the system enables memory resources to be used for multiple purposes while maintaining IP prefix matching accuracy.
Data Source
AI summary
a content addressable memory circuit is provided that includes a memory array that includes multiple memory devices that include memory locations that share a memory address and are coupled for simultaneous access. Hash logic is operative to use modulo math to determine a memory address based upon non-X values within an IP address key. Memory controller logic is operative to cause a memory device in the memory array to store the received IP address key in a memory location at the determined memory address, in a format that includes a field-size value indicative of a number of non-X values within a received IP address key and that includes non-X values within the received IP address key.