Flow-Based Congestion Notification for Network Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network congestion notification methods are inefficient as they require end-to-end communication, leading to delayed congestion awareness at the source device, which can result in increased buffer sizes and packet drops.
Innovation Solution
A network device generates a congestion notification packet (CNP) that is directly sent to the source device, indicating specific packet flow congestion, allowing for immediate flow-based rate control and reducing latency in congestion response.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If end-to-end congestion notification is used, then the source device can adjust transmission rate based on congestion feedback, but the congestion awareness is delayed due to the need for returning packets to traverse the entire path back to source
Solution Approach 1:
The patent segments the congestion notification process by introducing intermediate network devices (routers/switches) that can independently generate and send congestion notifications directly to source devices. This breaks the traditional end-to-end notification chain into localized segments where each network device monitors and reports congestion independently, eliminating the delay caused by packet round-trips through the entire network path.
Solution Approach 2:
The patent introduces intermediate network devices as mediators between the congested network path and the source device. These intermediary devices monitor traffic flow, detect congestion conditions, and directly communicate congestion status to source devices through dedicated notification packets, bypassing the need for acknowledgment packets to travel back through the entire network path.
2Reliability
If larger buffers are used to handle congestion, then packet drops are reduced, but the device complexity and resource requirements increase
Solution Approach 1:
The patent implements preliminary action by having network devices proactively detect congestion conditions and send notification packets to source devices before buffers become full and packet drops occur. This advance warning allows source devices to reduce transmission rates preemptively, preventing buffer overflow situations and eliminating the need for large buffers to handle congestion aftermath.
Solution Approach 2:
The patent establishes a direct feedback mechanism where intermediate network devices continuously monitor traffic flow and immediately notify source devices of congestion conditions. This real-time feedback loop enables source devices to dynamically adjust transmission rates based on actual network conditions, maintaining packet delivery reliability without requiring oversized buffers as a static safety margin.
Data Source
AI summary
A network device includes first, second, and third processors. The first processor detects congestion in a packet flow. The packet flow is i) one packet flow among a plurality of packet flows and ii) is formed of a plurality of packets of a same type received from a first device in a network via a first network connection. The packets in the packet flow are destined for a second device in the network. When congestion notification packet generation is enabled for the packet flow, the second processor generates a congestion notification packet by replicating a packet from the packet flow and sends the congestion notification packet to the first device via the first network connection. The congestion notification packet identifies the packet flow for which congestion is detected. The third processor forwards the plurality of packets in the packet flow to the second device via a second network connection.