Hash Content Addressable Memory Segmentation for Die Area Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ternary Content Addressable Memory (TCAM) devices require large die area and high power consumption due to the large number of transistors needed to store one bit of data, which results in significant heat dissipation and inefficiencies in packet forwarding and classification operations.
Innovation Solution
The Hash Content Addressable Memory (HCAM) system uses a hash function to calculate the location of stored search values, allowing for simultaneous comparison and reducing the number of transistors required by distributing table entries across multiple hash CAM blocks based on shared masks, thereby minimizing power consumption and die area.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If TCAM cells use two SRAM cells and comparison logic to store one bit of data, then the storage capability is achieved, but the die area per storage bit becomes large
Solution Approach 1:
The patent segments the TCAM into multiple blocks, each containing multiple segments. Each segment stores a portion of the key bits, allowing the comparison operation to be distributed across segments. This segmentation reduces the area required per storage bit by organizing the storage structure more efficiently.
Solution Approach 2:
The patent introduces a new dimension of organization by dividing keys into multiple segments stored in different segments within blocks. Instead of storing complete keys in single rows, the system uses a multi-dimensional structure where keys are split across segments (e.g., segment 0, segment 1, segment 2), reducing the area per bit while maintaining storage capability.
2Productivity
If all compare logic transistors in all TCAM cells are activated during search, then the search operation is performed, but the power consumption and heat dissipation become large
Solution Approach 1:
The patent divides the TCAM into multiple blocks with multiple segments each. During search operations, only the necessary segments and blocks are activated based on the search key, rather than activating all compare logic transistors across the entire TCAM. This segmented approach reduces power consumption while maintaining search functionality.
Solution Approach 2:
The patent implements partial action by activating only the necessary segments and blocks during search operations. Instead of fully activating all compare logic, the system selectively enables only the portions needed for the current search, reducing overall power consumption and heat dissipation while still achieving the search objective.
3Speed
If TCAM is organized as an array with multiple rows for simultaneous comparison, then the packet forwarding speed is improved, but the device complexity and die area increase
Solution Approach 1:
The patent segments the TCAM into multiple blocks, each containing multiple segments. This segmentation allows simultaneous comparison operations to be distributed across segments while reducing the complexity within each individual segment. The hierarchical structure (blocks containing segments) maintains high-speed parallel search capability while lowering overall device complexity.
Solution Approach 2:
The patent introduces a hierarchical dimension with blocks containing multiple segments, transforming the flat array structure into a multi-level organization. This dimensional change allows parallel search operations to occur across segments while reducing the complexity of individual comparison units, achieving speed improvement without proportional complexity increase.
Data Source
AI summary
Output is produced from a memory block. A key memory is accessed to produce memory output based on a value of a first index. The memory output includes a stored compressed key and a stored index. A compressed key is produced based on the uncompressed key. The produced compressed key is compared with the stored compressed key. The stored index is output when the stored compressed key matches the produced compressed key.