Selective Backpressure for Network Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In packet flow management, destination nodes can run out of buffer space, leading to packet dropping, and applying backpressure to upstream flows can cause congestion and slow other unrelated flows, as existing methods lack effective differentiation between node states and timely resource utilization.
Innovation Solution
A multi-state framework is introduced, assigning destination nodes as good actor, bad actor, or backpressure states, where network elements adjust behavior based on these states, limiting backpressure to only undesirable nodes, and hardware drops packets destined for bad actor states to prevent memory overflow.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If backpressure is applied to destination nodes with insufficient buffer space, then packet dropping is prevented, but congestion is caused to other unrelated flows
Solution Approach 1:
The patent applies backpressure selectively to specific destination nodes based on their individual buffer status rather than applying it globally to all flows. Each destination node is evaluated independently, and backpressure is applied only to nodes that are truly in need (those with insufficient buffer space), allowing other unrelated flows to continue without congestion.
Solution Approach 2:
The patent changes the state parameter of destination nodes from a binary state to a multi-state system (good actor state, bad actor state, backpressure state). This allows the system to differentiate between nodes based on their buffer utilization patterns and apply appropriate actions, transforming how backpressure is managed from a uniform approach to a differentiated approach.
2Reliability
If backpressure is applied to all destination nodes, then packet dropping is reduced, but network congestion increases
Solution Approach 1:
The patent extracts the backpressure mechanism from a global application and applies it only to specific destination nodes that require it. By identifying and isolating the nodes with buffer issues (bad actor states), the system applies backpressure only where necessary, preventing the harmful effect of congestion on unrelated flows while still protecting against packet dropping at problematic nodes.
3Productivity
If destination nodes are differentiated into good actor and bad actor states, then backpressure application is optimized, but system complexity increases
Solution Approach 1:
The patent implements a dynamic state management system where destination nodes can transition between good actor state, bad actor state, and backpressure state based on real-time buffer conditions. This dynamic approach allows the system to adapt to changing network conditions automatically, optimizing backpressure application without requiring complex manual configuration or intervention.
Data Source
AI summary
In an embodiment, a method includes, in response to detecting available memory of a destination node of a packet flow of nodes to the destination node being below a particular threshold, marking the destination node as being in a backpressure state. The destination node, in the backpressure state, sends a signal indicating a condition of packet backpressure to the nodes of the packet flow, and initiates a timer for a particular time period. The method further marks, at the end of the particular time period, the destination node as being in a bad actor state if the available memory is below the particular threshold, and as being in a good actor state if the memory is above the particular threshold. The method, in response to marking the destination node as being in a bad actor state, sends a signal to the nodes causing the nodes to drop packets directed to the destination node.