The How and When of Cisco FabricPath
Since we announced FabricPath at CiscoLive!, the most common questions I have gotten are “how do I get it?” and “when should I use it?”
So, the first question is pretty easy to answer. While we launched the Fabric Switching System (FSS) as part of the announcement as a pre-built option, the reality is that all you need is one or more of the new F-series IO modules to able to deploy FabricPath. This would be an example of the forward investment protection we talked about when we first introduced the Nexus 7000–customers have an simple way to add FabricPath support to their existing N7K chassis in a incremental fashion. Availability is calendar Q3 of this year.
The “when should I use it” question is a little bit more interesting because there is no hard and fast answer.
When we introduced FabricPath, we talked about a couple of common use cases: 1) support for applications that are deployed in a federated server environment and 2) virtualized server environments. These two environments will benefit from the large bi-sectional bandwidth, deterministic latency and any-vlan-anywhere features of FabricPath; however, these are not the only scenarios that could take advantage of the features (in the press release both NASA and Lawrence Livermore National Lab mentioned other use cases). So the first part of the “when” answer is to deploy FabricPath opportunistically–where you will gain some benefit from it. This point is key. The reality is that the typical enterprise had less than 20% of production workloads virtualized and only select applications running in a federated server environment. So, while FabricPath is incredibility useful, it is also something that you want to deploy selectively–it does not make a whole lot of sense to rip out your existing architecture and replace it with a collapsed FabricPath architecture if there is no underlying tangible benefit. The second part of the “when” answer is that your answer will change over time. As your percentage of virtualized workloads grows and as you roll out new business systems based on new application architectures, we expect FabricPath to propagate in the data center; however, I expect the typical data center to always remain a blend of STP/vPC and FabricPath, but the ratios will change over time. The nice thing about the Nexus is that adjusting to those changing ratios is a simple swap of IO modules.