Bigger Edge Workloads: Multi-Chassis X-Direct with Nutanix

In my last post, we talked about the UCS X-Series and the development / launch of the X-Direct / Fabric Interconnect Modules (FIM / integrated FIs). In that post, I purposefully left information out which answers a question often asked by customers: "What if I need to expand beyond 8 modular servers?"

Legitimate question to which the answer has historically been something along the lines of; If a customer anticipates going beyond 8 modular servers in a single site, they should consider doing the full UCS X-Series with IFMs (Intelligent Fabric Modules) connected to FIs (Fabric Interconnects).

Fortunately, Cisco's answer has evolved. If a customer wants to go down a full UCS X-Series deployment at an edge site, they can absolutely do so, however, Cisco recently introduced support for Multi-Chassis X-Direct deployments, which is now officially qualified and supported with Nutanix.

This architecture and solution, for some customers, can be the entire data center solution. Just because it's called a "Bigger Edge" doesn't mean that this can't be used at the core of a data center. Like I mentioned in the first post, 12 different definitions from 10 different engineers for what an "edge site" is.

Let's dig in.

Multi-Chassis X-Direct

First lets address why I'm making a "big" deal out of this: A Cisco UCS Nutanix cluster is bound to a single UCS domain (or FI pair). This means that in the X-Series world, the following would not be supported.

This implies that with the X-Direct, if a customer tried to expand their 8 node cluster to a 9 node (or more) cluster with a new chassis / nodes, it would not work as the integrated FIs (the Fabric Interconnect Modules [FIM]) in the chassis come up against the "single UCS domain" limitation. Similar to the video above but considering the X-Direct, below is an example of what WAS not supported. And it's worth clarifying that, technically speaking, the image below is still not supported. Keep reading to see the difference with the connectivity and what the supported solution entails.

Until now!!! Cisco recently (albeit very quietly) released support for multi-chassis X-Direct support with Nutanix. This allows customers to scale at the edge with up to two (2) UCS X-Series chassis (1 X-Series Direct and the other with regular IFMs) and when fully populated, up to 16 modular servers. Not only that, but it also includes the ability to add up to four (4) C-Series rackmount servers. The solution provides expanded use case support for the large edge sites (or smaller data centers).

Supported Architectures with X-Direct & Multi-Chassis Support:

• HCI / HCI + CO / CO + External Storage
  • Up to 16 X-Series nodes in a single cluster
  • Add up to 4 rackmount servers for a total of 20
  • Multiple clusters supported if required
  • For HCI + CO - read the architecture limitations and ratio scaling
  • Ability to still mix Nutanix and non-Nutanix related workloads within the same architecture
• A mixed setup within the same site:
  • Compute Only nodes + External Storage as a cluster
  • HyperConverged Infrastructure (HCI) nodes as a cluster
  • HyperConverged Infrastructure (HCI) + Compute Only (CO) nodes

That's a lot of possibilities. Let's take a look at a few of these.

Use Cases

First, let's show what a fully maxed out configuration looks like. I mentioned this in my last post but it is worth repeating here: The image below does not need to be all dedicated to Nutanix. Nutanix software is merely a workload that consumes resources. Running Nutanix next to Windows next to Red Hat is fully supported.

Next, I find it somewhat important to show that this solution does support CO + External Storage; FlashStack with Nutanix (and soon to be FlexPod with Nutanix):

The above image shows that an Everpure storage array can be directly connected to Chassis #1 via both IP and FC protocol and serve storage to appropriate workloads using the appropriate protocol. In the case of Nutanix, this is NVMeoF/TCP.

And last but not least, the everything model, including the kitchen sink of possibilities:

Any workload, any chassis, single UCS domain - Full flexibility.

There are of course more iterations of this that could be shown but hopefully the images/videos above start to pain a picture of all the possible permutations.

So let's take a look at what Cisco offers in conjunction with Nutanix:

• Cisco UCS X-Series for the core data center or extra large edge sites
• Cisco UCS X-Series Direct (X-Direct) multi-chassis support + up to 4 rackmount servers for smaller data centers or the large edge sites
• Cisco UCS X-Series Direct (X-Direct) for the small to medium edge sites
• <purposefully left blank>

Does a bullet that says it was purposefully left blank actually blank? One can debate, but the real question you should be asking yourself is: "What about the extra small to small sites???"