Adaptive Congestion Control and Load Balancing for Network Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing congestion control mechanisms in computer networks result in inefficient bandwidth utilization and increased overhead due to frequent congestion notifications and suboptimal path selection, leading to network latency and reduced performance in high-performance applications like AI and HPC.
Innovation Solution
Implementing congestion control and load balancing mechanisms that adjust congestion notification frequency based on congestion levels, recycle entropies for efficient path selection, leverage RTT and ECN signals for congestion detection, and optimize packet delivery contexts to minimize overhead and improve bandwidth utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If congestion control mechanisms send frequent congestion notifications to respond to network congestion, then congestion response promptness is improved, but network overhead increases and bandwidth utilization becomes inefficient
Solution Approach 1:
The patent applies dynamics by making the congestion notification frequency adaptive rather than static. The sending rate limit is dynamically adjusted based on the congestion level metric, allowing the system to send notifications more frequently when congestion is severe and less frequently when the network is healthy, thus resolving the contradiction between response promptness and overhead
Solution Approach 2:
The patent changes the parameter of notification frequency based on the congestion level metric. By adjusting the sending rate limit parameter according to measured congestion conditions, the system optimizes the balance between timely congestion detection and minimizing network overhead, directly addressing the technical contradiction
2Productivity
If load balancing mechanisms frequently change network paths to optimize traffic distribution, then bandwidth utilization is improved, but path selection overhead increases and network latency increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and caching path selection metrics before congestion occurs. The system prepares path selection information in advance based on current network conditions, so when congestion is detected, the load balancer can quickly switch to pre-evaluated paths without incurring significant latency, thus resolving the contradiction between bandwidth optimization and latency
3Reliability
If congestion control mechanisms reduce sending rate aggressively to alleviate congestion, then congestion relief is improved, but bandwidth utilization becomes inefficient
Solution Approach 1:
The patent applies dynamics by making the sending rate reduction adaptive to the measured congestion level. Rather than applying fixed aggressive rate limits, the system dynamically adjusts the reduction magnitude based on the congestion metric, allowing for gentle rate adjustments when congestion is mild and more significant reductions only when necessary, thus resolving the contradiction between congestion relief and bandwidth utilization efficiency
Data Source
AI summary
A computing network implements a congestion control mechanism and a load balancing mechanism. It is determined which available routes in the network are congested. Activation of the congestion control mechanism is limited until a threshold number of the available routes are determined to be congested, which prevents over-attenuation of an overall sending rate or window by the congestion control mechanism.