OpenFlow Match-Action Pipeline Multicast Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current hardware switch architectures face challenges in efficiently processing multicast packets due to the need for high-speed output of multiple packet copies to multiple destinations, which increases complexity and cost, making it difficult to optimize the match-action pipeline for total input bandwidth.
Innovation Solution
The implementation of OpenFlow ALL groups in the egress processor, where a single copy of a packet is stored in a common data buffer and modified by each bucket, allowing for efficient output of multiple copies to different ports without requiring dedicated structures, thereby simplifying the design and sharing resources with existing match-action structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If dedicated structures are used for each packet copy in multicast processing, then high-speed output to multiple destinations is achieved, but device complexity and cost increase
Solution Approach 1:
The patent merges the packet storage function into a shared common data buffer that serves all multicast destinations. Instead of having dedicated buffer structures for each output port, a single buffer stores the original packet and generates multiple copies through software processing, thereby reducing hardware complexity while maintaining high-speed output capability
Solution Approach 2:
The common data buffer is designed to serve multiple functions: storing unicast packets, storing multicast packets for replication, and supporting the match-action pipeline. This universal structure eliminates the need for separate dedicated structures for each packet type and destination, reducing overall device complexity
2Reliability
If dedicated structures are used for each packet copy in multicast processing, then reliable delivery to multiple destinations is ensured, but manufacturing cost increases
Solution Approach 1:
The patent combines multiple packet handling functions into a single common data buffer structure, reducing the total amount of hardware required. This merging approach lowers manufacturing costs while maintaining reliable delivery through software-based packet replication and modification for each destination
Solution Approach 2:
The system creates software copies of the packet from a single stored instance in the common buffer. Each copy is generated and modified as needed for specific destinations, ensuring reliable delivery without requiring expensive dedicated hardware structures for each packet instance
3Productivity
If the match-action pipeline is optimized for total input bandwidth, then processing efficiency is improved, but multicast packet handling becomes more difficult
Solution Approach 1:
The packet is stored in the common data buffer before multicast replication occurs. This preliminary storage action allows the match-action pipeline to process the original packet efficiently while the buffer handles the replication task, separating the optimization concerns and allowing the pipeline to focus on bandwidth optimization without complicating multicast handling
Data Source
AI summary
An embodiment of the invention includes a packet processing pipeline. The packet processing pipeline includes match and action stages. Each match and action stage in incurs a match delay when match processing occurs and each match and action stage incurs an action delay when action processing occurs. A transport delay occurs between successive match and action stages when data is transferred from a first match and action stage to a second match and action stage.