Cisco FabricPath – Scaling your data center network
Recently, I was having a discussion with some customers about how data center networks are being redefined by convergence and virtualization, and how Cisco FabricPath is helping IT staff scale their data center networks to make them more agile for virtualization deployments. If you’re not familiar with Cisco FabricPath technology, it’s a superset of the IETF TRILL standard with added features. Cisco FabricPath allows you to build out highly scalable, multi-path layer 2 networks without Spanning-Tree.
In typical server virtualization deployments, VLAN scalability is confined to a rack or pod due to the design constraints of Spanning-Tree. Cisco FabricPath eliminates this constraint allowing you to extend VLANs across all racks in the data center. Essentially any switch port in your data center can be in any VLAN. This is especially beneficial in a virtualized data center since it allows seamless workload mobility across the data center. Check out this FLASH for a high level overview of how FabricPath works.
Back to the customer conversation – several questions about FabricPath /TRILL architecture came up. Some of these questions have come up before, so I figured I’d share them with a wider audience.
Q1. Will Cisco support TRILL
A1. Cisco supports FabricPath today and will support TRILL in the near future giving customers a choice which one to deploy. Customers implementing FabricPath today and wanting to switch over to TRILL will be able to do so with a simple software upgrade.
Q2. With FabricPath, will I have to implement a multi-tier architecture for Layer 3, multicast and FCoE?
A2. No, Cisco’s Nexus series switches support all these functions simultaneously provide a scalable layer 2 data center network while also providing Layer 3, multicast and FCoE functionality . No additional equipment layers are required to support these additional functions.
Q3. With large FabricPath broadcast domains, will broadcast storms ,flapping interfaces or malicious applications cause widespread disruptions ?
A3. No, FabricPath does allow you to build out large VLANs or broadcast domains that can stretch across the data center. However, unlike spanning tree, broadcast packets are handled very differently. This is because for all practical purposes, FabricPath is “routing” traffic inside the Fabric using IS-IS in the background. This means that a link flap will be handled with IS-IS and will have not result in flooding traffic or a broadcast storm.
Q4. Does FabricPath provide multi-pathing for inter-VLAN traffic ?
A4. Yes, Cisco FabricPath provides multiple active default gateways, thus allowing multi-pathing for inter-VLAN routing (one of those added features I mentioned earlier). Cisco is working with the IETF to add this FabricPath capability to TRILL.
Q5. How many VLANs can FabricPath support?
A5. FabricPath and TRILL currently offers 4K VLANs, as defined in the 802.1Q standard. Additional segmentation can be achieved over TRILL using conventional mechanisms available today.
Q6. With FabricPath, will we be able to implement congestion management schemes like Qcn?
A6. Yes, Fabricpath does not preclude the use of Qcn. Congestion management can be viewed as a function that is platform specific so there is no reason why a FabricPath/TRILL based system could not implement a Qcn mechanism. For example, Cisco platforms implement Data Center Bridging IEEE protocols, a standard way of providing a lossless Ethernet fabric. This standard mechanism runs on Ethernet, irrespective of whether the network is running TRILL.
For more information on FabricPath and how it works, visit…. http://www.cisco.com/en/US/prod/switches/ps9441/fabric_path_promo.html
We hope to see you soon at Cisco Live ( hash tag #cldc11) in a couple of weeks , please don’t hesitate to stop at our booth for more details.