Dual-Homed Active-Active Devices for L2 Broadcast Loop Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dual-homed, active-active endpoints in network fabrics cause loops in Layer-2 broadcast messages, leading to issues such as incorrect endpoint identifiers and endpoint identifier flaps due to the endpoints receiving and forwarding messages from multiple edge nodes, resulting in inefficient network operations.
Innovation Solution
Edge nodes in the network fabric identify the source MAC address of Layer-2 broadcast messages, storing it in a local map-cache and comparing it with existing addresses to determine if the message has been looped. If matched, the message is dropped; otherwise, it is added to a source MAC address table and registered with the control-plane, ensuring efficient message handling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If dual-homed endpoints are connected to multiple edge nodes in an active-active fashion, then performance and redundancy are improved, but message loops and incorrect endpoint identifiers occur
Solution Approach 1:
The patent applies preliminary action by pre-registering endpoint identifiers with a specific edge node before messages are forwarded. When an endpoint is dual-homed, its identifier is initially registered with one edge node (the primary), and this registration is stored in a database. This preliminary registration prevents message loops because other edge nodes can check the database before forwarding messages, determining whether the message should be dropped to avoid loops. This resolves the contradiction by maintaining redundancy (multiple active connections) while preventing harmful message loops through pre-established registration information.
2Productivity
If dual-homed endpoints forward messages from multiple edge nodes, then load balancing is achieved, but endpoint identifier flaps and network inefficiency occur
Solution Approach 1:
The patent implements feedback by having edge nodes check a centralized database before forwarding messages to dual-homed endpoints. When an edge node receives a message, it queries the database to determine if the endpoint identifier is registered with that specific edge node. Based on this feedback from the database, the edge node decides whether to forward the message or drop it to prevent loops. This feedback mechanism enables load balancing across multiple active connections while preventing endpoint identifier flaps, because the database provides authoritative information about which edge node should receive messages for each endpoint.
3Reliability
If edge nodes forward all L2 broadcast messages to dual-homed endpoints, then complete message delivery is ensured, but redundant message forwarding increases network traffic
Solution Approach 1:
The patent extracts the harmful redundant message forwarding by separating the message forwarding decision from the standard L2 broadcast behavior. Instead of automatically forwarding all L2 broadcast messages to dual-homed endpoints (which causes redundancy and energy waste), the system extracts the decision-making process and bases it on database lookups. The edge node queries the database to determine if the endpoint is registered with that node, and only forwards messages when appropriate. This extraction eliminates redundant message forwarding while ensuring complete message delivery through registered paths, resolving the contradiction between reliability and energy efficiency.
Data Source
AI summary
Techniques for edge nodes of a network fabric to identify L2 broadcast messages that have been looped through connections with dual-homed, active-active endpoints. The edge nodes receive an L2 broadcast message from the network fabric, and store a source MAC address of the message in a local map-cache. The edge devices then forward the L2 broadcast message to the endpoint. The edge nodes may receive the messages that have been looped through the connections with the endpoint, and compare the source MAC address of the messages with the source MAC addresses stored in the locasl map-cache. If the source MAC address of the message matches to an address in the local map-cache, the edge nodes drop the message. However, if the source MAC address does not match to an address in the local map-cache, the edge nodes add the source MAC address to a source MAC address table.