Network Congestion Control via Flow Predictability and Bandwidth Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data center networks face congestion issues due to throughput-intensive loads, leading to high delays and degraded throughput, especially when delay-sensitive applications share networks with throughput-intensive applications, resulting in network bottlenecks and packet loss, which existing solutions address only reactively and do not prevent congestion effectively.
Innovation Solution
A method and apparatus that classify flows as predictable or non-predictable based on data transmission patterns, using a network controller to allocate bandwidth and schedule data packet transmission, employing hybrid genetic algorithms and simulated annealing to optimize congestion transmission models and minimize queue lengths in switches, thereby preventing congestion proactively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If throughput-intensive applications occupy large cache space in switches, then data transmission capacity is improved, but delay-sensitive applications suffer from high delay and packet loss
Solution Approach 1:
The patent segments network flows into throughput-intensive flows and delay-sensitive flows, applying different handling strategies to each segment. Throughput-intensive flows are allowed to occupy cache space while delay-sensitive flows are prioritized for immediate transmission, resolving the contradiction between transmission capacity and delay performance
Solution Approach 2:
The patent applies different quality of service (QoS) characteristics to different regions of the network buffer. Certain buffer regions are designated for delay-sensitive traffic with strict priority, while other regions handle throughput-intensive traffic, ensuring that local buffer management optimizes both transmission capacity and delay performance
2Adaptability or versatility
If multiple connections contend for the same egress buffer, then network connectivity is improved, but buffer overflow occurs causing packet loss
Solution Approach 1:
The patent implements preliminary congestion detection mechanisms that monitor buffer utilization before overflow occurs. When congestion thresholds are approached, the system proactively applies congestion control measures such as packet dropping or transmission rate limiting, preventing buffer overflow and maintaining reliable packet delivery while preserving network connectivity
Solution Approach 2:
The patent implements feedback mechanisms where congestion information is continuously monitored and fed back to transmitting nodes. This feedback enables dynamic adjustment of transmission rates and routing decisions, allowing multiple connections to share buffer resources efficiently while preventing overflow and maintaining reliable delivery
3Reliability
If TCP retransmission is required due to packet loss, then data integrity is improved, but network throughput drastically decreases
Solution Approach 1:
The patent implements prior cushioning by proactively detecting congestion conditions and applying preventive measures before packet loss occurs. Through early congestion detection and reactive packet dropping, the system prevents the need for TCP retransmissions, thereby maintaining both data integrity and high network throughput without the performance penalty of retransmission timeouts
Solution Approach 2:
The patent converts the potentially harmful effect of packet loss into a beneficial congestion signal. By intentionally dropping packets under congestion conditions and using this as feedback for flow control, the system achieves efficient throughput management while maintaining data integrity, turning what would normally require costly retransmissions into a useful mechanism for congestion avoidance
4Reliability
If existing congestion control solutions are applied, then congestion response is improved, but proactive congestion prevention cannot be achieved
Solution Approach 1:
The patent implements preliminary congestion detection mechanisms that monitor network conditions and identify congestion patterns before they fully develop. By detecting early signs of congestion such as increasing queue lengths or changing traffic patterns, the system can proactively apply control measures to prevent congestion, achieving both reliable response and proactive prevention
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of the present invention disclose a method and an apparatus for handling network congestion, and a system. The method includes: obtaining a transmitted data volume of a flow, and identifying a predictable flow and a non-predictable flow in the flow; collecting statistics about total data transmission volumes of the predictable flow and the non-predictable flow; obtaining a congestion transmission model of the predictable flow, and solving the congestion transmission model to obtain a scheduling policy for the predictable flow; and allocating bandwidths to the predictable flow and the non-predictable flow to obtain a bandwidth allocation result, and sending the bandwidth allocation result and the scheduling policy to the host, so that the host executes the scheduling policy in a scheduling period. This can prevent congestion in advance and reduce a delay of a delay-sensitive flow, and is applicable to a large data center.