Serial Packet Replication Engine for Network Switch Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The replication time of packets in network switches contributes significantly to packet processing latency, and existing parallel replication methods consume excessive processing resources and face circuit layout restrictions.
Innovation Solution
The implementation of serial packet replication, where packets are replicated over sequential clock cycles using a replication engine that analyzes replication information and packet pointers stored in a work FIFO, allowing for efficient distribution across multiple output ports while minimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If parallel replication is used to reduce packet processing latency, then replication speed is improved, but processing resources and circuit layout complexity increase excessively
Solution Approach 1:
The patent segments the packet replication process into sequential stages across multiple clock cycles rather than executing all replications simultaneously. Each clock cycle handles a portion of the replication task, dividing the workload to reduce resource requirements while maintaining acceptable performance.
Solution Approach 2:
The patent implements dynamic packet scheduling where the replication engine can adaptively determine which packets to replicate in each clock cycle based on current system state and queue conditions. This dynamic approach allows flexible resource allocation without requiring fixed parallel replication infrastructure.
2Loss of time
If parallel replication is implemented to minimize replication time, then packet processing latency is reduced, but the number of processing resources required increases significantly
Solution Approach 1:
The patent uses a work queue to pre-schedule packet replication tasks in advance. Packets are staged in the work queue and processed in sequential batches, allowing the system to prepare replication tasks beforehand without requiring multiple simultaneous replication engines, thus reducing the quantity of processing resources needed.
Solution Approach 2:
The patent implements periodic replication processing where packets are replicated in periodic intervals across multiple clock cycles rather than continuously in parallel. This periodic approach maintains replication speed while significantly reducing the peak resource consumption required for simultaneous processing.
Data Source
AI summary
Disclosed are various embodiments that provide serial replication of multicast packets by performing a first data fetch to fetch first data from a memory buffer, the first data comprising a first packet pointer representing a first packet and a replication number indication a number of times the first packet is to be replicated. Furthermore, various embodiments are directed to performing a second data fetch to fetch second data from a memory buffer, the second data comprising a first packet pointer representing a second packet and serially replicating the first packet and the second packet based at least in part upon the replication number and a predetermined threshold value.