Packet Classification Using Relevance Scoring Hierarchical Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern network devices face challenges in quickly and efficiently processing protocol data units (PDUs) due to the large number of policies required for traffic classification, which burdens the system and affects transfer rates in gigabit networks.
Innovation Solution
A multi-stage classification method that uses hierarchical tables to match strings with policies, generating scores for each match and selecting the best rule based on cumulative scores, allowing for efficient processing and reducing the burden on network devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple strings (thousands of policies) are extracted from PDUs for traffic classification, then classification accuracy and policy coverage are improved, but processing time and device burden increase significantly
Solution Approach 1:
The patent divides the classification task into multiple stages by segmenting the policy set into hierarchical tables. Each table contains a subset of policies organized by criteria, allowing the system to process policies in stages rather than evaluating all thousands of policies simultaneously. This segmentation reduces the processing burden at each stage while maintaining comprehensive policy coverage.
Solution Approach 2:
The patent introduces a hierarchical dimension to the classification process by organizing policies into multiple levels of tables. Instead of a flat single-stage classification, the system uses a multi-dimensional hierarchical structure where each level filters policies based on specific criteria. This dimensional organization enables efficient navigation through the policy space and reduces processing time.
2Adaptability or versatility
If a large number of policies (thousands) are maintained for traffic classification, then classification completeness is improved, but device complexity and processing burden increase
Solution Approach 1:
The patent segments the large policy set into manageable hierarchical tables, each containing a subset of policies. This segmentation allows the system to maintain comprehensive classification capabilities while reducing the complexity of individual processing stages. Each table can be processed independently and efficiently.
Solution Approach 2:
The patent performs preliminary organization of policies into hierarchical tables with predefined criteria before actual classification occurs. This preliminary structuring enables the system to navigate and process policies more efficiently during runtime, reducing the processing burden while maintaining completeness.
3Adaptability or versatility
If traditional single-stage classification is used with thousands of policies, then policy coverage is maintained, but transfer rates in gigabit networks cannot be maintained
Solution Approach 1:
The patent segments the classification process into multiple stages that can operate in parallel or pipeline fashion. By dividing the policy evaluation into stages, the system maintains high transfer rates in gigabit networks while preserving comprehensive policy coverage. The hierarchical structure enables efficient processing that does not bottleneck at high speeds.
Solution Approach 2:
The patent introduces a hierarchical dimension to classification that enables scalable processing. This multi-level structure allows the system to maintain productivity at gigabit rates by organizing policy evaluation in a way that does not linearly scale with the number of policies, thus preserving both policy coverage and transfer rate capability.
Data Source
AI summary
A method and apparatus for performing multi-stage classification of a plurality of strings extracted from a packet, for example, is disclosed. The method causes a plurality of hierarchical tables to be searched, each table having criteria derived from the policies with which the packet is classified. One of the plurality of the strings is searched at each of the plurality of tables. Each match identified in one table is used to retrieve a pointer to a subsequent table. Each of the matches in each of the hierarchical tables is scored in order to generate a cumulative score for each policy, a score may be used to select from a plurality of applicable policies. In a first embodiment, a plurality of policies, as embodied in the hierarchical tables, are searched and the best policy selected. In a second embodiment, recursive searching is employed.