Virtualization in the data center today enables us to pack more servers and applications into smaller spaces than ever before. Cloud technologies and mobility mean we might want to be able to easily connect these data centers with remote data centers as well. Unfortunately, these revolutionary changes are too much for classic technologies like Spanning Tree Protocol to handle.
The Internet Engineering Task Force (IETF) has been hard at work on Transparent Inter-Connection of Lots of Links (TRILL) as a replacement technology for STP. Cisco takes TRILL and enhances it to give us FabricPath in networks today.
In the video below, I provide an example of configuring Cisco FabricPath with Nexus switches.
FabricPath overcomes the limitations of Spanning Tree Protocol (STP)-based networks. STP-based networks possess these shortcomings:
- While we typically build redundancy into the important links of the data center, STP cannot take advantage of them. The protocol blocks redundant links, instead of permitting the advantages such as load sharing of multiple paths in the data center.
- Spanning Tree Protocol is not great at choosing the best path in the data center. The election of a root switch and other legacy mechanisms in STP often lead to poor path selections.
- Because STP operates in a pure Layer 2 environment, it lacks protection features such as the Layer 3 Time To Live (TTL) value, which are needed in order to guard against network-crashing storms in the network.
- MAC address scalability becomes a huge concern as every switch learns the MAC address of every device in the data center, and potentially beyond the local data center.
With TRILL, the idea is to take the simplicity and flexibility of Layer 2, and enhance it with the intelligence of Layer 3 -- specifically, the intelligence of a scalable, link-state routing protocol. IS-IS is the chosen protocol. Yes, it’s an exciting comeback for a protocol that lost out decades ago to OSPF.
Based on TRILL, Cisco FabricPath is exciting for a number of reasons. Here's a partial list:
- FabricPath is a simple configuration. While IS-IS routing intelligence is key to the operation of the technology, engineers are shielded from any raw configuration of the Layer 3 protocol. It just works, nestled inside the Layer 2 operations. New troubleshooting capabilities now present themselves at Layer 2, including ping and traceroute, to make problem solving all the simpler.
- FabricPath loves multiple paths in the data center right out of the box. No more wasting links and bandwidth due to STP. Need to add bandwidth or redundancy? Not an issue. Just add links.
- Because FabricPath is based on Layer 3 routing intelligence, it is excellent at choosing preferred paths in the network topology.
- Finally, our Layer 2 data center technology helps to reduce the MAC address table sizes on our switches. Conversational MAC learning in the FabricPath solution enables selective learning of the MAC addresses based on active flows, which can help us dramatically reduce table sizes in the data centers we manage today.
While many organizations right now might be contemplating a leap to Rapid Spanning Tree Protocol, others are dumping STP altogether and reaping the rewards of FabricPath.