IP Performance Measurement for Faster EVPN Multihoming Convergence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Ethernet Virtual Private Network (EVPN) systems experience slow convergence and oversubscription due to delays in notifying transmitting Provider Edge (PE) devices of failures at Ethernet segments (ES), particularly in multihoming scenarios.
Innovation Solution
Implementing an IP performance measurement session between Provider Edge devices to rapidly convey health state information using a field in the message, enabling failover actions when connection failures occur, such as through a Two-Way Active Measurement Protocol (TWAMP) or Bi-directional Forwarding Detection (BFD), to reroute traffic through alternate paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If BGP is used for signaling in EVPN, then routing control is achieved, but convergence speed is slow due to delays in notifying transmitting PE devices of Ethernet segment failures
Solution Approach 1:
The patent implements performance measurement sessions that continuously monitor Ethernet segment health and provide feedback about failure conditions to transmitting PE devices. This feedback mechanism enables rapid detection and notification of failures, eliminating the delays inherent in traditional BGP-based failure notification.
Solution Approach 2:
The patent establishes performance measurement sessions in advance before failures occur, allowing transmitting PE devices to be pre-configured with monitoring mechanisms. When failures happen, these pre-established sessions immediately detect and report the conditions, enabling faster convergence without waiting for BGP updates.
2Reliability
If BGP delay notification is used for Ethernet segment failures, then network stability is maintained, but traffic oversubscription occurs during convergence
Solution Approach 1:
Performance measurement sessions provide real-time feedback on Ethernet segment health, allowing transmitting PE devices to detect failures immediately and adjust traffic routing before oversubscription occurs. This continuous monitoring prevents the harmful effects of delayed failure notification while maintaining network stability.
Solution Approach 2:
By establishing performance measurement sessions beforehand, the network is prepared to detect and respond to failures immediately. This preliminary setup prevents traffic oversubscription by enabling proactive failure detection and rapid rerouting, eliminating the need to choose between stability and preventing harmful traffic patterns.
3Device complexity
If traditional BGP failure notification is implemented, then protocol simplicity is maintained, but convergence time increases
Solution Approach 1:
The patent leverages existing performance measurement infrastructure and protocols (such as BFD and TWAMP) to achieve fast failure detection. These universal protocols can be integrated into the EVPN framework without requiring entirely new complex mechanisms, thus reducing convergence time while limiting the increase in overall system complexity.
Solution Approach 2:
The patent introduces performance measurement sessions as intermediary components between Ethernet segments and BGP signaling. These sessions act as mediators that provide rapid failure detection and notification, enabling faster convergence without fundamentally altering the BGP protocol itself, thus balancing complexity and performance improvements.
Data Source
AI summary
A system and method are provided for faster convergence with multihoming using internet protocol (IP) performance measurement (PM) sessions to report, to a transmitting provider edge (PE), a state of health of a connection between a multihomed node and a receiving (PE). The state of health can be the liveness of the connection. The receiving PE locally monitors the state of health of its connection to the node, and encodes information of the state of health in a field of the PM session messages sent to the transmitting PE. For example, while the PM session messages indicate the connection is live, traffic from the transmitting PE to the node is routed through the receiving PE. Upon the PM session messages indicating the connection is no longer live, the transmitting PE instead routes traffic to the node through another receiving PE with which the node is multihomed.