SCI Packet Classifier Memory Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current packet classification algorithms face scalability issues due to rapid rule set growth, leading to increased memory consumption and construction time, particularly in large rule sets, and are hindered by rule replication, which reduces performance in high-end routers, switches, and firewalls.
Innovation Solution
The Split, Compression, and Intersection (SCI) method, which splits rule sets based on overlapping frequencies, compresses them into Indexed Shape Graphs, and uses binary search trees to reduce memory footprint and enhance lookup efficiency, allowing for parallel processing and efficient rule matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If decomposition or decision-tree algorithms are used for packet classification, then packet classification capability is achieved, but memory consumption and construction time grow rapidly with rule set size
Solution Approach 1:
The patent merges multiple rule sets into a unified data structure called a rule cube, where overlapping rules are consolidated rather than replicated. This combining approach eliminates redundant rule copies across different decomposition branches, significantly reducing memory consumption while maintaining complete packet classification coverage.
Solution Approach 2:
The patent implements a nested structure where rule cubes are organized in hierarchical levels, with each level containing compressed representations of rules from lower levels. This nesting allows the system to store rules in a compact form where higher-level cubes reference and compress data from lower-level cubes, reducing overall memory requirements.
2Productivity
If decomposition or decision-tree algorithms are used for packet classification, then packet classification capability is achieved, but construction time grows rapidly with rule set size
Solution Approach 1:
The patent performs preliminary actions by pre-processing the rule set during the construction phase to identify and consolidate overlapping rules into the rule cube structure. By organizing rules into this optimized data structure in advance, the system enables faster online packet classification without repeatedly processing the entire rule set during each classification operation.
Solution Approach 2:
The patent changes the organizational parameters of rule storage from traditional flat or tree structures to a multi-dimensional rule cube structure. This parameter change in how rules are indexed and stored allows for more efficient construction algorithms that can exploit dimensional relationships, reducing construction time for large rule sets.
3Speed
If TCAM hardware is used for packet classification, then packet classification is processed in one clock cycle, but the system becomes inflexible and power hungry
Solution Approach 1:
The patent replaces the specialized TCAM hardware mechanism with a software-based algorithmic solution using rule cubes. This substitution maintains high classification speed through optimized data structures and algorithms while providing the flexibility and adaptability of software, allowing the system to handle dynamic rule updates and various classification requirements that TCAM cannot accommodate.
Data Source
AI summary
A SCI (Split, Compression and Intersection) packet classifier, wherein the SCI packet classifier comprises a construction module to generate rule ID array, h pairs of ISG and three binary search trees, and a lookup engine to look up the binary search trees and ISGs to find the highest priority rule matching the incoming packet and to output a corresponding rule ID.