Avatar

Authored by Karishma Gupta: Technical marketing Engineer, Data Center Networking, Cisco

What is Fabric Builder

Fabric Builder is a turn-key function within Cisco Data Center Network Manager (DCNM) version 11 for easy Day-0 provisioning of data center network deployments.

Why do you need it

With Fabric Builder, we allow users to quickly build a VXLAN EVPN fabric with minimal user input. The templates provided contain all the necessary functions to define the underlay, the VTEPs and the overlay control-plane, while incorporating best practices to build the fabric in the most efficient way. Templates used in Fabric Builder can be customized, and new templates can be created using the flexible Python+ infrastructure in DCNM with ample flexibility for a user to accommodate any configuration across any subset of devices. These “freeform” configurations can be created at an interface/device(s)/fabric level. Fabric Builder allows Day 0-2 provisioning for all Nexus platforms – be it for LAN Fabric (VXLAN EVPN based fabrics) or Classic LAN deployments.

How it works

Within DCNM 11, a user can get immediately started by navigating to Control-> Fabric Builder to create a fabric; there is no need to download or install any auxiliary programs or templates. The fabric creation starts by simply entering a name for the fabric and the BGP Autonomous System (AS) number. Default fabric settings can be kept as is or amended while creating the fabric. Various combinations are supported including numbered/unnumbered IP addressing for the core interfaces, routing protocols like OSPF and ISIS, replication modes like multicast (BiDir and ASM) and ingress replication.

Once the fabric is created, switches can be imported by optionally entering a seed switch with an appropriate hop count to allow shallow discovery of other connected switches. Fabric Builder allows for both, greenfield and brownfield deployments – in the case of existing configured switches, they can be imported into the fabric while preserving all the current configurations and state on the switches if the user desires to do so. For the case of new switch deployments, they can be bootstrapped by using POAP. Post addition, all discovered switches become part of the fabric which can be managed from DCNM.

At this stage, all the user needs to do is instruct DCNM on the switch roles, setup vPC domains with a one-click vPC pairing option and the entire fabric configuration is constructed based on the entered fabric settings and the discovered links between all the switches. As a network device can have various responsibilities, various switch roles exist, this includes roles like Leaf, Spine, Border leaf, Border at Spine, Border Gateway, and Border Gateway at Spine.

Now, DCNM’s Fabric Builder is ready to push the generated configurations to the switches. Prior to this, Fabric Builder provides a side-by-side comparison between the user defined intent and running configurations in the switch (in case there is one). Post validation, the configurations can be deployed at a fabric/switch/interface level. Voila! This completes the setup of a full Data Center fabric through DCNM 11’s Fabric Builder, in a few minutes. Incremental configuration and changes can be easily pushed to the switches via the Fabric builder throughout the life cycle of the fabric. While all the operations described above are possible via the GUI, all of them are equally available via REST-APIs.

If you have more than one data center

DCNM 11’s Fabric Builder allows the user to create multiple fabrics and also a construct called Multi-Site domain (MSD). MSD allows to hierarchically sort multiple individual fabrics and automatically configure VXLAN EVPN Multi-Site for Data Center Interconnect (DCI). Furthermore, MSD allows creation of Networks and VRFs only once, these are then inherited by each member fabric – more about this is in a follow-up article about DCNM 11’s Multi-Site capabilities.

Greenfield provisioning for VXLAN underlay using Fabric builder

 



Authors

Venu Venugopal

VP of Engineering

Intent Based Networking Group