Multi-Connection Congestion Control for Mixed-RTT Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing TCP congestion control algorithms struggle to find the optimal balance between round-trip time and latency in diverse network environments, particularly when transitioning between high-speed, low-latency networks and high-latency networks like LEO connections.
Innovation Solution
A network device dynamically switches between different congestion control algorithms for multiple connections based on performance criteria, replacing existing connections with new ones if certain conditions are met, and aggregates connections to optimize data packet transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single TCP congestion control algorithm is used, then the system is simple to implement, but it cannot adapt to diverse network environments with different latency characteristics
Solution Approach 1:
The patent implements dynamic switching between different congestion control algorithms based on real-time network conditions. The system monitors network performance metrics and transitions from one algorithm to another when conditions change, allowing the congestion control mechanism to adapt dynamically to diverse network environments including high-speed low-latency networks and high-latency networks like LEO connections
Solution Approach 2:
The system changes the parameter of congestion control algorithm selection based on network conditions. By monitoring parameters such as round-trip time and network performance, the system selects appropriate algorithms (e.g., BBR for high-speed networks, Cubic for traditional networks) and switches between them when performance deteriorates, optimizing performance for each network type
2Speed
If TCP congestion control uses slow start stage to establish cwnd baseline, then it can infer bandwidth availability, but it cannot quickly adapt to high-latency networks with significant RTT
Solution Approach 1:
The system uses feedback mechanisms to monitor network performance in real-time and adjust congestion control algorithms accordingly. By continuously measuring metrics like throughput, packet loss, and RTT, the system receives feedback about network conditions and switches algorithms to optimize performance for high-latency networks while maintaining high transmission speeds
3Reliability
If TCP congestion control quickly reduces cwnd when congestion is detected, then it avoids overloading the network, but it reduces data transfer efficiency during transient congestion
Solution Approach 1:
The system changes the congestion control parameters and algorithm selection based on the type and duration of congestion detected. For transient congestion in high-speed networks, the system may switch to algorithms like BBR that maintain higher throughput while avoiding overload. For persistent congestion, it switches to more conservative algorithms, optimizing both reliability and productivity based on actual network conditions
Data Source
AI summary
The present invention discloses a system and method for performing congestion control at an apparatus. The apparatus may establish at least one first connection with a first node and establish a plurality of second connections with a second node. After the at least one first connection and the plurality of second connections are established, the apparatus may transmit data packets to the second node through the plurality of second connections according to an outbound traffic policy when receiving the data packets from the first node through the at least one first connection. If a first criterion is satisfied, the apparatus may establish a plurality of third connections with the second node and transmit data packets to the second node through the plurality of third connections when receiving data packets from the first node through the at least one first connection.