Ethernet Congestion Control via MAC Address Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ethernet networks lack the ability to differentiate between congestive and non-congestive data flows, leading to inefficient flow control and resource utilization, as back-pressure mechanisms throttle all traffic upstream of a congested node without distinguishing between problematic and acceptable flows.
Innovation Solution
A method that detects congestion, sends congestion messages upstream, learns MAC address information of affected nodes, and broadcasts this information to control data flow based on learned addresses, allowing for intelligent differentiation between congestive and non-congestive flows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If back-pressure mechanism is used to control congestion, then congestion at the congested node is reduced, but all upstream traffic is throttled including non-congestive flows
Solution Approach 1:
The patent segments the upstream traffic into different flows based on destination MAC addresses. Instead of applying back-pressure uniformly to all upstream traffic, the congestion control mechanism identifies and selectively applies flow control only to specific destination addresses that are causing congestion, while allowing other non-congestive flows to continue uninterrupted.
Solution Approach 2:
The patent implements local quality by making the flow control action selective rather than uniform. The congestion control is applied locally to specific destination MAC addresses identified as problematic, rather than globally to all upstream traffic. This allows different treatment for different flows based on their contribution to congestion.
2Loss of information
If Ethernet bridge learns every MAC address passing through, then address information is available for forwarding, but inability to determine hop distance prevents congestion source identification
Solution Approach 1:
The patent performs preliminary action by having upstream nodes learn and store destination MAC addresses before congestion occurs. When congestion is detected, the pre-learned address information is immediately available for rapid identification of affected flows, eliminating the need for complex real-time analysis of hop distances or congestion sources.
Data Source
AI summary
A method of controlling data flow in a network includes detecting a congestion condition in the network, sending a congestion message to one or more nodes upstream of the congestion condition, learning address information of one or more nodes pertaining to the congestion condition, sending another congestion message containing these learned MAC addresses and, via the nodes receiving the congestion message with the MAC addresses, controlling a flow of data associated with the learned address information from the congestion message. The learned address information may be the MAC address of a destination node to which the data is flowing towards or a MAC address pair of the nodes between which the data is flowing.