VXLAN MAC Table Scaling via FID Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing technologies face scalability issues in handling Border Gateway Protocol (BGP) route information in Virtual eXtensible Local Area Networks (VXLAN) with Ethernet Virtual Private Network (EVPN) control plane, particularly in data plane operations and control plane advertisements.

Innovation Solution

The proposed solution involves programming the Media Access Control (MAC) address table using Filtering Identifiers (FIDs) instead of Bridge Domains (BDs), and introducing a new EVPN NLRI Route Type for VTEP's Router MAC routes to reduce redundant advertisements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If BGP route information is handled using traditional Bridge Domain (BD) based MAC address table programming in VXLAN with EVPN control plane, then routing functionality is provided, but the number of MAC address table entries becomes excessively large in large-scale deployments

Engineering Contradiction:
Improverouting scalabilityVSAvoidnumber of MAC address table entries
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

The patent segments the traditional Bridge Domain (BD) concept into two separate components: Bridge Domain (BD) for layer 2 broadcasting and Filtering Identifier (FID) for layer 3 routing. This segmentation allows MAC address table entries to be programmed with FIDs instead of full BD identifiers, significantly reducing the number of bits required per entry and enabling scalable deployment in large-scale VXLAN networks.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a new dimensional organization for MAC address table entries by adding FID as a separate dimension from BD. Instead of using a single BD identifier that grows in size with network scale, the solution uses a fixed-size FID dimension combined with a BD dimension, creating a multi-dimensional indexing structure that scales efficiently.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Reliability

If router MAC advertisements are sent for every VRF and Bridge Domain combination in EVPN control plane, then reachability information is distributed, but redundant advertisements increase control plane overhead

Engineering Contradiction:
Improvereachability information distributionVSAvoidcontrol plane overhead
Core Design Contradiction:
ReliabilityVSLoss of energy

Solution Approach 1:

The patent extracts the FID component from the traditional BD-based advertisement mechanism. By separating FID from BD in the advertisement structure, the system can advertise router MAC reachability information more efficiently, sending advertisements only when necessary rather than for every VRF-BD combination, thereby reducing control plane overhead while maintaining reliable reachability distribution.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter structure of EVPN advertisements by introducing FID as a distinct parameter from BD. This parameter change allows the advertisement mechanism to be more selective and efficient, reducing the frequency and volume of redundant advertisements while ensuring that necessary reachability information is properly distributed across the EVPN control plane.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3975487B1Scalable handling of BGP route information in vxlan with EVPN control plane
Publication Date: 2025.01.29 CISCO TECHNOLOGY INC
  • EP3975487B1 patent drawingFigure 1
  • EP3975487B1 patent drawingFigure 2
  • EP3975487B1 patent drawingFigure 3

AI summary

One aspect relates to the issue of how to efficiently utilize route information for the support of a large scale deployment of VXLAN overlay with EVPN control plane. A method for programming a MAC address table (250) by a first leaf node in a network (210) comprising a plurality of leaf nodes (210-1, 220-2, 220-3, 220-4) is provided. Each leaf node comprises one or more Virtual Tunnel End Points, VTEPs, and instantiates a plurality of Virtual Routing and Forwarding elements, VRFs, with a corresponding Bridge Domain, BD, assigned to each VRF. The method includes obtaining information indicating one or more VTEP Affinity Groups, VAGs, each VAG comprising an identification of one VTEP per leaf node, obtaining information indicating assignment of each VRF to one of the VAGs, assigning each VAG to a unique Filtering Identifier, FID, thereby generating one or more FIDs, and programming the MAC address table (250), using FIDs instead of BDs, by populating the MAC address table with a plurality of entries, each entry comprising a unique combination of a FID and a MAC address of a leaf node. As a result of the assignment of each VAG to a unique FID, core BDs of the VRFs which are assigned the same VAG are mapped (270) into the FID assigned to the VAG. The VTEP affinity approach to scale MAC address table programming in a data path provides enhancements to existing handling of BGP route information in VXLAN with EVPN control plane. As a result, the number of MAC address table entries needed for a large scale deployment of VXLAN with EVPN as control plane may be substantially reduced. Another aspect relates to the issue of how to cut down the redundant control path advertisements for better control path scalability. A new route type (VTEP's Router MAC route) for the EVPN NLRI to scale router MAC advertisement in control path provides enhancements to existing handling of BGP route information in VXLAN with EVPN control plane. As a result, the number of router MAC advertisements needed for a large scale deployment of VXLAN with EVPN as control plane may be substantially reduced.