TCP Congestion Control via User Feature-Based Window Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional TCP congestion control methods are inefficient, especially in scenarios with multiple terminals accessing the network, leading to reduced resources and poor user experience due to equal treatment of all flows, which does not account for varying network conditions and user behaviors.
Innovation Solution
A TCP congestion control method that determines a recommended congestion window based on key features of user flows, using a model established through clustering and deep learning algorithms to adjust the congestion window proactively, avoiding network congestion by distributing different transmission resources to different users based on their specific conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional TCP congestion control algorithms (such as Newreno, cubic, westwood) are used, then TCP protocol compatibility is maintained, but network resource utilization is insufficient and congestion window growth is too slow
Solution Approach 1:
The patent changes the parameters of congestion window growth by introducing multiple growth rates (first growth rate, second growth rate, third growth rate) based on different network conditions and user types. Instead of using a fixed or standard TCP growth pattern, the system dynamically adjusts the congestion window expansion speed to optimize network resource utilization while maintaining protocol compatibility.
Solution Approach 2:
The patent implements dynamic congestion window adjustment by switching between different growth rates according to real-time network status (congested vs. non-congested) and user flow characteristics. The congestion window is not statically determined but dynamically adapted during transmission, allowing the system to respond to changing network conditions and maximize resource utilization.
2Ease of operation
If conventional TCP congestion control treats all flows equally, then protocol fairness is maintained, but effective user resources become insufficient and user experience deteriorates
Solution Approach 1:
The patent applies local quality by differentiating treatment for different user flows based on their characteristics (effective users vs. other users). Instead of uniform congestion window allocation, the system assigns different growth rates and resource priorities to different user types, allowing effective users to obtain more network resources while maintaining basic fairness for all users.
Solution Approach 2:
The patent segments the user base into different categories (effective users identified by specific criteria vs. other users) and applies different congestion control strategies to each segment. This segmentation allows the system to optimize resources for high-priority users while still providing service to all users, resolving the contradiction between fairness and resource sufficiency.
3Productivity
If congestion window is increased rapidly to maximize bandwidth utilization, then network throughput improves, but network congestion occurs and stability decreases
Solution Approach 1:
The patent applies preliminary action by detecting network congestion status in advance and adjusting the congestion window growth rate accordingly. When congestion is detected, the system switches to a slower growth rate or reduces the window size before severe congestion occurs, preventing network instability while still allowing rapid throughput increase when the network is clear.
4Ease of operation
If conventional TCP determination methods (RTT time delay, packet loss) are used to detect network congestion, then detection simplicity is maintained, but detection accuracy is insufficient and one-sided
Solution Approach 1:
The patent implements feedback mechanisms by continuously monitoring multiple parameters (RTT variations, packet loss patterns, throughput changes) and using this feedback to adjust congestion window decisions. The system incorporates acknowledgment feedback from the receiver and adjusts transmission based on the combination of these feedback signals, improving detection accuracy beyond simple RTT or packet loss thresholds.
Data Source
Figure 1~2
Figure 3-1
Figure 3-2
AI summary
The present application provides a TCP congestion control method, an apparatus, a terminal and a readable storage medium. The TCP congestion control method includes acquiring key features of a user, acquiring an ISP resource generated according to a TCP flow request message of the user and caching the acquired ISP resource into a cache queue, determining a recommended congestion window of a TCP protocol layer control block according to the key feature, and then transmitting a data packet of the cache queue to the user according to a principle of sliding window.