Multi-Stage Network Switch Congestion Root Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-stage network switches face challenges in efficiently detecting and addressing sustained incast congestion due to the cost and complexity of equipping each stage with congestion detection mechanisms, and shallow buffers that struggle to distinguish between temporary and sustained congestion.
Innovation Solution
Implementing a system where only the output stage of a multi-stage network switch includes an incast congestion detection mechanism, with upstream stages operating at a higher packet transfer rate to fill output buffers quickly, allowing for accurate determination of sustained congestion by monitoring buffer fullness and credit management between stages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detection mechanisms are equipped at each stage of the multi-stage network switch, then congestion detection accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The patent extracts the congestion detection function from all stages and concentrates it only in the output stage. The output stage independently determines whether it is the root of congestion by monitoring its own buffer conditions and credit status, eliminating the need for detection mechanisms at upstream stages while maintaining accurate congestion detection capability
Solution Approach 2:
The output stage performs multiple functions: it acts as both the data transmission node and the congestion detection node. By making the output stage universal in handling both data forwarding and congestion determination, the patent eliminates redundant detection mechanisms at other stages, reducing overall device complexity while preserving detection accuracy
2Device complexity
If shallow buffers are used in multi-stage network switches, then device complexity is reduced, but the ability to distinguish between temporary and sustained congestion deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the output stage continuously monitors its buffer fullness and credit status over time. By observing whether congestion conditions persist across multiple credit cycles, the system can distinguish between temporary congestion (which resolves quickly) and sustained congestion (which persists), enabling accurate congestion type differentiation despite using shallow buffers
Solution Approach 2:
The output stage proactively monitors buffer conditions and credit status in advance to detect congestion patterns. By tracking the duration and persistence of buffer fullness conditions before making congestion determination, the system can differentiate congestion types without requiring large buffers to store additional measurement data
3Reliability
If detection mechanisms are added at each stage, then congestion detection capability is improved, but manufacturing cost increases
Solution Approach 1:
The patent extracts the congestion detection functionality from multiple stages and consolidates it into a single output stage. This reduction in the number of detection mechanisms directly lowers component count and manufacturing complexity, reducing production costs while maintaining the capability to accurately identify the root of congestion in the network fabric
4Stability of the object's composition
If upstream stages operate at lower packet transfer rates, then buffer overflow is prevented, but network throughput decreases
Solution Approach 1:
The patent implements dynamic packet transfer rates at upstream stages based on real-time credit availability from the output stage. When credits are available, upstream stages operate at higher rates to maximize throughput; when credits are depleted, rates are reduced to prevent buffer overflow. This dynamic adjustment allows the system to achieve both high throughput and buffer stability adaptively
Solution Approach 2:
The credit-based mechanism ensures continuous useful action by maintaining a steady flow of packets from upstream stages to the output stage. By continuously granting credits when buffer space is available and continuously monitoring buffer status, the system maintains optimal throughput without interruption while preventing buffer overflow, ensuring uninterrupted data transmission
Data Source
AI summary
Technologies for determining a root of congestion include a network switch. The network switch is to operate arbiter units in at least one upstream stage at a packet transfer rate that is greater than a packet transfer rate of an arbiter unit in an output stage, determine whether an input buffer of a remote network switch in communication with the output stage has sustained congestion over a first predefined time period, determine whether an output buffer of the arbiter unit in the output stage has sustained congestion over a second predefined time period, and determine, as a function of whether the input buffer of the remote network switch has sustained congestion and whether the output buffer of the arbiter unit in the output stage has sustained congestion, whether the network switch is a root of congestion.