Port Priority Flow Control for FCoE Switches
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Packet loss due to switch buffer overflow caused by traffic congestion in packet-switched networks, particularly in Ethernet networks carrying Fiber Channel over Ethernet (FCoE) traffic, which cannot tolerate substantial packet loss, is not effectively addressed by existing priority-based flow control (PFC) systems that lead to head-of-line (HOL) blocking and inefficient traffic control.
Innovation Solution
The implementation of a priority- and port-based flow control (PPFC) system that associates each end system facing port with specific queues at an aggregation/core switch, allowing only the congested queue to be blocked while allowing other queues to service non-congested ports, using a hierarchical scheduling architecture and FCoE Initialization Protocol (FIP) for queue establishment and backpressure requests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If priority-based flow control (PFC) is implemented to manage traffic congestion, then traffic prioritization is improved, but head-of-line (HOL) blocking occurs causing inefficiency
Solution Approach 1:
The patent segments the flow control mechanism by introducing port identification to PFC, dividing the network into port-specific flow control domains. This allows independent flow control per port while maintaining priority levels, preventing HOL blocking by enabling selective backpressure application to specific ports rather than blocking all high-priority traffic across the network.
Solution Approach 2:
The patent applies local quality by making flow control properties port-specific rather than uniform across all ports. Each port can have its own backpressure state independent of other ports, allowing localized traffic management that prevents HOL blocking while maintaining overall network productivity through differentiated flow control characteristics.
2Reliability
If existing PFC systems block traffic to prevent buffer overflow, then packet loss is reduced, but all queues are blocked including non-congested ports
Solution Approach 1:
The patent segments the blocking mechanism by introducing port-specific identification to flow control. When buffer overflow is detected on a specific port, only that port's traffic is blocked while other ports continue normal operation. This segmentation prevents unnecessary blocking of non-congested ports while maintaining reliability through targeted backpressure application.
Solution Approach 2:
The patent applies local quality by making flow control blocking port-specific rather than network-wide. Each port can independently enter and exit blocked states based on its own congestion conditions, allowing non-congested ports to maintain full productivity while congested ports receive protective blocking to prevent packet loss.
3Reliability
If traditional flow control blocks all high-priority traffic, then buffer overflow is prevented, but traffic granularity and control precision are reduced
Solution Approach 1:
The patent segments flow control into port-specific control units, enabling precise identification and management of individual ports within high-priority traffic. This segmentation allows the system to track congestion states per port rather than treating all high-priority traffic uniformly, thereby maintaining buffer overflow prevention while significantly improving control precision through port-level granularity.
Solution Approach 2:
The patent applies local quality by assigning unique flow control properties to each port. The port identification field enables differentiated treatment of individual ports, allowing precise measurement and control of congestion states at the port level rather than aggregate level, thus improving traffic control granularity while maintaining reliability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An apparatus comprising an aggregation/core switch configure to couple to an edge switch and receive information about a plurality of end system facing ports of the edge switch, wherein the information about the end system facing ports is used to associate the end system facing ports with a plurality of corresponding queues at the aggregation/core switch. Also disclosed is a network component comprising a receiver configured to receive information about a plurality of end system facing ports of an edge switch, a processor configured to establish and associate the end system facing ports with a plurality of corresponding queues, and a transmitter configured to return information about the associated end system facing ports.