Distributed Switch Partition Handling via Local State Retention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High availability solutions in FC/FCoE networks face issues due to synchronization challenges between primary and secondary fibre channel over Ethernet (FCoE) forwarders, leading to potential service outages when link failures cause system partitioning, resulting in loss of traffic and service disruptions.
Innovation Solution
Implementing a method that maintains local and global connection state information between primary and secondary FCoE forwarders, using an alternate address mechanism and tunneling between controlling FCoE forwarders and end switching hops to ensure continuous service by routing frames to the correct partition, thereby avoiding loss of traffic and service outages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If connection state is synchronized from primary FCF to secondary FCF, then high availability is achieved, but system partitioning causes loss of synchronization and global state
Solution Approach 1:
The patent segments the distributed switch into multiple independent FCF nodes, each maintaining its own local connection state. This segmentation allows each node to operate autonomously during partitions, preventing total system failure while maintaining high availability through distributed state management rather than centralized synchronization.
Solution Approach 2:
Each FCF node maintains local connection state information locally rather than relying on centralized synchronization from the primary FCF. This local quality approach ensures that each node has the necessary state information to continue operating independently during system partitions, resolving the synchronization loss problem while maintaining reliability.
2Productivity
If traffic is routed during system partition, then service continuity is maintained, but traffic may be black holed and not reach destination
Solution Approach 1:
The patent implements dynamic routing where FCF nodes adaptively adjust traffic forwarding based on real-time partition detection. When a partition is detected, nodes dynamically modify their forwarding behavior to route traffic through available paths while preventing black hole routing, thereby maintaining service continuity without causing traffic loss.
Solution Approach 2:
The system employs feedback mechanisms where FCF nodes continuously monitor system state and partition conditions. This feedback enables nodes to make informed routing decisions, dynamically adjusting traffic flow to avoid black holes while maintaining service continuity, thus resolving the contradiction between productivity and harmful factors.
3Device complexity
If primary/secondary FCF model is used, then operational control is simplified, but all connection intelligence is concentrated in primary FCF creating single point of failure
Solution Approach 1:
The patent segments the control intelligence from the primary FCF and distributes it across multiple FCF nodes. Each node maintains local connection state and can independently make forwarding decisions, eliminating the single point of failure while preserving operational control simplicity through standardized node behavior and partition detection protocols.
Solution Approach 2:
Each FCF node is designed to be universal and multi-functional, capable of operating as both primary and secondary depending on partition conditions. This universality allows any node to assume control responsibilities, eliminating the single point of failure inherent in dedicated primary/secondary models while maintaining simplified operational control through consistent node functionality.
Data Source
AI summary
An example of a distributed system partition can include a method for client service in a distributed switch. The method can include maintaining local and global connection state information between a primary and a secondary controlling fiber channel (FC) over Ethernet (FCoE) Forwarders (FCFs) or FC forwarder in a distributed switch. A partition in the distributed switch can be detected and service to subtended clients of the distributed switch can continued using local state information.