Network Flow Congestion Management via Intermediate Node Marking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for managing network congestion are inadequate due to high latency, coarse-grained congestion notification, uncoordinated resource utilization, and lack of precise congestion location awareness, leading to suboptimal resource utilization and quality of service issues.
Innovation Solution
A system and method that utilize a central management entity to collect and analyze congestion information from multiple sources, marking packets with the highest congestion level along a flow path, allowing for coordinated and targeted congestion mitigation strategies, including flow grouping and resource allocation adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If end-to-end explicit congestion notification is used, then congestion can be detected and notified to source nodes, but higher end-to-end latency causes delay in taking control measures
Solution Approach 1:
The patent introduces intermediate network nodes (routers, switches, or controllers) as mediators that perform local congestion detection and notification. These intermediaries monitor queue utilization and resource usage at network nodes and send congestion notifications directly to source nodes or flow endpoints, bypassing the need for end-to-end detection. This reduces latency by eliminating the round-trip time required for end-to-end congestion detection while maintaining accurate congestion information through local monitoring.
2Ease of operation
If coarse-grained congestion notification is used, then congestion can be indicated to source nodes, but no precise information about congestion level and location is available
Solution Approach 1:
The patent segments congestion information into multiple granular components including congestion level (queue utilization percentage), location (specific network node identifier), resource type (CPU, memory, bandwidth), and flow identifier. Instead of providing a single binary congestion indication, the system divides congestion data into discrete, precise elements that can be independently analyzed and acted upon by source nodes or controllers.
Solution Approach 2:
The patent uses packet marking with different congestion levels (e.g., ECN codepoints indicating mild, moderate, severe congestion) to provide precise congestion information. This is analogous to using different colors to indicate different states, where the marking carries rich information about congestion severity and type without requiring complex signaling protocols.
3Productivity
If physical scaling of network resources is performed, then resource bottlenecks can be relieved, but the exponential increase in demand outpaces the average revenue per user
Solution Approach 1:
The patent implements dynamic resource allocation and flow routing based on real-time congestion feedback. Source nodes or controllers adjust flow rates, prioritize traffic, or reroute flows dynamically in response to congestion notifications, maximizing the utilization of existing network resources without requiring physical expansion. This dynamic adaptation allows the network to handle variable demand efficiently.
Solution Approach 2:
The patent changes operational parameters (flow rates, routing paths, priority levels, buffer sizes) in response to congestion conditions rather than changing physical infrastructure. By adjusting these parameters dynamically, the system optimizes network performance and capacity utilization within existing resource constraints, avoiding the need for proportional physical scaling.
4Productivity
If multiple resource types are virtualized and consolidated, then resource utilization efficiency improves, but resource bottlenecks caused by processing load and I/O interfaces increase
Solution Approach 1:
The patent implements comprehensive feedback mechanisms that monitor not only network queue utilization but also processing load, memory usage, and I/O interface status at virtualized network functions. This multi-dimensional feedback enables the system to identify bottlenecks in processing resources and adjust flow routing or rates accordingly, preventing congestion caused by computational limitations while maintaining high resource utilization.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method for managing data flows in a network with a plurality of forwarding elements routing flows between network entities (UD, N, SN, FE) of a network or network domain (ND), wherein a) flows between network entities are marked with congestion information, preferably in form of resource utilization information, by at least one of the forwarding elements such that the highest congestion level prevailing on the respective flow path is indicated, b) the locations of one or more possible and/or present bottlenecks and/or congestions in the network based on the congestion information are identified, and c) one or more actions are performed to avoid, mitigate, and/or to resolve the identified possible and/or present bottlenecks and/or congestions.