Distributed Congestion Avoidance in Network Switching Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network switching systems face congestion issues due to mismatched port speeds and oversubscription, where the offered load exceeds the processing or forwarding capacity, leading to inefficiencies and potential buffer overflow.
Innovation Solution
The implementation of a network switching system that includes multiple ingress devices with virtual output queues (VOQs) that exchange queue state information to adjust transmission rates, ensuring that the total offered traffic load does not exceed the capacity of shared resources by distributing the load fairly across multiple VOQs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If FIFO buffers are used to absorb excess traffic near congestion points, then traffic buffering capacity is improved, but system complexity and potential buffer overflow increase
Solution Approach 1:
The system performs preliminary congestion detection by monitoring queue state information before congestion occurs. Ingress devices exchange queue state messages to anticipate congestion conditions and adjust traffic loads proactively, rather than reactively buffering overflow traffic.
Solution Approach 2:
The system implements feedback mechanisms where ingress devices continuously exchange queue state information about VOQ fullness with other devices. This feedback enables dynamic adjustment of transmission rates to maintain traffic load within capacity limits, preventing buffer overflow without requiring large FIFO buffers.
2Productivity
If transmission rates are increased to improve throughput, then productivity is improved, but congestion occurs when offered load exceeds processing capacity
Solution Approach 1:
The system dynamically adjusts transmission rates based on real-time queue state information. Ingress devices modify their transmission rates adaptively according to the fullness of destination VOQs, allowing throughput to vary within safe limits rather than operating at a fixed high rate that could cause congestion.
Solution Approach 2:
Transmission rate adjustments are driven by feedback from queue state messages exchanged between ingress devices. The system continuously monitors VOQ states and uses this information to modulate transmission rates, ensuring productivity is maximized without exceeding processing capacity and causing congestion.
3Difficulty of detecting and measuring
If queue state information is exchanged between ingress devices, then congestion detection capability is improved, but communication overhead increases
Solution Approach 1:
The system extracts only the essential congestion-related information (VOQ fullness state) from the complete queue state and exchanges this condensed information between ingress devices. This selective information exchange provides sufficient congestion detection capability while minimizing communication overhead.
Solution Approach 2:
The system changes the parameter representation of queue state from detailed metrics to simplified fullness indicators (e.g., thresholds of VOQ occupancy). This parameter transformation reduces the amount of information that must be exchanged while maintaining the ability to detect congestion conditions accurately.
Data Source
AI summary
According to an example embodiment, a total offered traffic load for a shared resource within a network switching system may be determined, the total offered traffic load may include, for example, a sum of offered traffic loads from one or more active virtual output queues (VOQs) of the network switching system. A capacity of the shared resource within the network switching system may be determined. A transmission rate from one or more of the active VOQs over the shared resource may be adjusted such that the total traffic load from the active VOQs does not exceed the capacity of the shared resource.