Redundant Ethernet Switch Fabric with Backpressure Congestion Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Chassis-based communication devices require advanced redundancy and congestion management features to achieve high availability, but standard networking technologies like Ethernet lack these features, relying on higher layer protocols and suffering from head of line blocking and data loss.
Innovation Solution
Implementing a switch fabric using standard Ethernet devices with redundancy logic to replicate and sequence frames, and employing virtual output queues with backward congestion notification to manage congestion and redundancy, allowing for near-lossless data transfer and high throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If standard networking technologies like Ethernet are used for switch fabric, then cost is reduced and standard interfaces are provided, but redundancy features and congestion management capabilities are lost
Solution Approach 1:
The system segments the redundancy function into separate components: the standard Ethernet switch fabric handles basic switching, while independent redundancy logic units (RLUs) at each endpoint handle frame replication and selection. This allows standard switches to be used while adding redundancy capabilities through software/firmware layers.
Solution Approach 2:
Redundancy logic units act as intermediary components between the standard Ethernet switch fabric and the application layers. These RLUs insert sequence numbers into frames, replicate frames to multiple switch modules, and monitor incoming frames to select which copies to forward, thereby adding redundancy functionality to standard networking infrastructure.
2Device complexity
If standard Ethernet switch fabric is used, then device complexity is reduced, but head of line blocking and data loss occur during congestion
Solution Approach 1:
The system performs preliminary actions by inserting sequence numbers into frames before they enter the switch fabric and setting up virtual output queues at endpoint devices. This preparation enables the system to handle congestion and recover from data loss without requiring complex real-time interventions during traffic bursts.
Solution Approach 2:
The system implements feedback mechanisms through backward congestion notification (BCN) frames that travel from congested endpoints back to source endpoints. When congestion is detected, BCN frames trigger flow control actions at the source, allowing the system to adapt to changing conditions and prevent data loss without requiring overly complex switch fabric architecture.
3Reliability
If purpose built switch fabric components are used, then redundancy and congestion management features are provided, but cost increases and standard interfaces are lost
Solution Approach 1:
The system achieves multi-functionality by combining standard Ethernet switch fabric with overlay redundancy logic. The same standard switch infrastructure serves both basic switching functions and, when combined with endpoint redundancy logic units, provides redundancy and congestion management capabilities that would otherwise require specialized hardware.
Solution Approach 2:
The system creates logical copies of frame transmission paths by replicating frames at the source endpoint through redundancy logic units and sending copies through the standard switch fabric via different switch modules. This copying approach enables redundancy without requiring duplicate specialized switch fabric hardware.
4Ease of operation
If Ethernet pause frames are used for congestion control, then implementation is simple, but head of line blocking occurs reducing throughput
Solution Approach 1:
The system applies local quality by implementing flow control at specific virtual output queues rather than globally. Each VOQ can be independently backpressured based on its own congestion state, allowing non-congested queues to continue transmitting at full rate while only affected queues are slowed down, thereby maintaining high overall throughput.
Solution Approach 2:
The system segments the flow control mechanism into queue-specific virtual output queues rather than using a single global pause frame approach. This segmentation allows differential flow control where only the specific queues experiencing congestion are backpressured, preventing head of line blocking from propagating to all traffic flows.
Data Source
AI summary
A communications apparatus includes several functional modules for implementing an application, a pair of switch modules to provide redundant switching capability for transferring frames between the functional modules. Each functional module is connected to each switch module by a frame-based communication link. A redundancy logic unit at each functional module inserts sequence numbers into outgoing frames, replicates the outgoing frames for forwarding to each of said switch modules, and monitors incoming frames from each of the switch modules to select frames for forwarding to an application based on the sequence numbers. In this way, redundancy is maintained at all times, while duplicate frames are eliminated at the receiving module.