Compound Mask Packet Classification Tree for Lookup Speed
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current packet classification systems face challenges in efficiently routing a variety of packets at high speeds due to high resource consumption and long lookup times, particularly with the use of TCAM hardware, which is expensive and power-intensive, and existing algorithms do not effectively manage different types of ACLs like IPv4, IPv6, and Ethernet.
Innovation Solution
The implementation of a packet classification tree using compound masks that traverse fewer steps and require fewer resources, leveraging entropy to select candidate bit masks and define bit tables, allowing for more efficient packet classification with reduced processing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If TCAM hardware accelerated algorithms are used for packet classification, then lookup speed is improved, but device cost and power consumption increase significantly
Solution Approach 1:
The patent segments the packet classification task into multiple stages, with each stage handling a subset of ACL types (e.g., IPv4, IPv6, Ethernet separately). This allows using simpler, cheaper hardware for each segment rather than requiring expensive TCAM for the entire classification task, thereby reducing overall device cost while maintaining high lookup speed through parallel stage processing.
Solution Approach 2:
The patent uses software-based packet classification algorithms that replicate the functionality of expensive TCAM hardware but run on general-purpose processors. This copying approach achieves comparable lookup performance without the high cost and power consumption of actual TCAM hardware, effectively substituting expensive hardware with software implementations.
2Loss of time
If TCAM hardware is used for packet classification, then lookup time is reduced, but memory resource consumption increases
Solution Approach 1:
The patent divides the classification rules into multiple stage-specific data structures, with each stage maintaining only the ACLs relevant to that stage. This segmentation reduces the memory footprint at each stage compared to storing all ACLs in a single TCAM, while still achieving fast lookup through the multi-stage parallel processing approach.
3Adaptability or versatility
If traditional packet classification algorithms are used to support multiple ACL types, then adaptability is improved, but processing complexity increases
Solution Approach 1:
The patent creates separate processing stages, each dedicated to a specific ACL type (IPv4, IPv6, Ethernet, etc.). This segmentation allows each stage to use optimized, simple algorithms tailored to its specific ACL type, rather than using a single complex algorithm that must handle all ACL types. The overall system achieves high adaptability through the combination of specialized stages while keeping individual stage complexity low.
Solution Approach 2:
The patent designs a universal multi-stage classification architecture that can handle multiple ACL types through a common framework. Each stage is designed to process specific ACL types, but the overall system provides universal support for various packet types and protocols, achieving both specialization and universality simultaneously.
Data Source
Figure 1~2A
Figure 2B
Figure 3~4
AI summary
A method in a network element of a packet data network classifies a data packet for application of a packet data network policy. A root node compound bit mask is applied to bits of a received data packet. The data packet has data in a plurality of fields. The compound bit mask has a plurality of bit masks each having bits in a different data packet field. The masked bits of the compound bit mask are applied to root node bit tables, one for each bit mask of the compound bit mask. A branch node from the root node is selected based on a bit table look up.