Explicit Congestion Signaling in Network Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current end-to-end congestion control algorithms (CCAs) rely on implicit and approximate signals for determining congestion, leading to suboptimal application performance and inefficient network usage, especially in scenarios requiring fast network transfers like AI/ML applications.
Innovation Solution
A network node is configured to enable explicit congestion control of network traffic using congestion signaling across various hosts executing CCAs, allowing for real-time communication of congestion signals to manage packets on a per-connection or per-packet basis, thereby optimizing network utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If end-to-end congestion control algorithms use implicit and approximate signals, then the algorithms can operate with simpler mechanisms, but the congestion control precision and network performance deteriorate
Solution Approach 1:
The patent introduces an intermediary congestion signal mechanism that operates between end hosts and network infrastructure. Network devices inject explicit congestion signals into data plane packets, which are then reflected back to senders. This intermediary layer provides precise congestion information without requiring complete redesign of end-to-end control algorithms, thus improving precision while maintaining reasonable complexity.
Solution Approach 2:
The patent replaces implicit, approximate congestion detection mechanisms with explicit, precise congestion signaling. Instead of relying on indirect indicators like packet loss or delay heuristics, the system substitutes these with direct congestion signals injected by network devices, providing accurate real-time congestion state information to congestion control algorithms.
2Loss of information
If explicit congestion signaling is implemented with separate packets, then congestion information can be communicated clearly, but the network overhead and packet size increase
Solution Approach 1:
The patent merges congestion signaling with existing data plane traffic by injecting explicit congestion signals directly into application data packets. Rather than creating separate control packets, the congestion information is embedded within the same packets that carry application data, utilizing unused header fields or payload spaces. This approach communicates complete congestion information without increasing overall network traffic volume.
3Stability of the object's composition
If congestion control operates on larger timescales with approximate signals, then the control system is more stable, but the response time to congestion changes slows down
Solution Approach 1:
The patent implements periodic congestion signal injection at network device boundaries, creating regular feedback intervals. Network devices periodically measure congestion metrics and inject corresponding signals into passing packets, providing senders with time-series congestion information. This periodic action maintains system stability through regular sampling while enabling fast response to congestion changes through the high frequency of signal updates.
Solution Approach 2:
The patent establishes a feedback loop where network devices continuously monitor congestion conditions and send explicit signals back to traffic sources. The congestion control algorithm receives this feedback and adjusts transmission rates accordingly. This closed-loop feedback mechanism enables both stability (through controlled adjustment) and fast response (through real-time congestion information).
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Congestion control by adding a congestion signal tag header to each of one or more transmission packets prior to transmission of the transmission packets by the first node to a second node, the congestion signal tag header specifying one or more congestion signal types and, for each of the congestion signal types, specifying a congestion signal value by providing an initial congestion signal value for the congestion signal value; receiving one or more return packets generated by the second node in response to receipt of the transmission packets, the return packets including a congestion signal reflection header having one or more return congestion signal values, and the return congestion signal values corresponding respectively to the congestion signal types; and determining whether transmission rate control is necessary based on the return congestion signal values.