Hyper Convergence – A new Era for Infrastructure

I’ve been working recently with Cisco’s HyperFlex System. If I could boil it all down to 1 word; AMAZING.

Why do I need HyperFlex you ask?

Key Benefits to HyperFlex

  • Ease of Setup – Cisco UCS is in my humble opinion one of the best Blade Servers ever made. It’s so easy to setup and use, but it is time consuming to configure. HyperFlex eliminates all this extra work by allowing you to Script the whole setup in just a few minutes. It’s truly unbelievable, and I was awestruck the 1st time I saw it run.
  • Complete Hyperconvergance – Fabric Network and Compute are unified in a Next Gen Platform. Spin ALL of that up in MINUTES!
  • Flash Optimized System – Flash Storage Built into each device. No need for External Storage unless you just wanna. The built in storage is designed for Hyperconvergance from the ground up!
  • Scaling – With HyperFlex you can simply plug in additional devices to scale the system.
  • Continuous Data Optimization – Data is continually optimized, de-duplicated and Compressed on the fly.

That’s why!

Setup Experience

The setup is so simple, although we had some initial confusion out of the gates on how the communications would be routed through ACI, That’s Cisco’s Application Centric Infrastructure. This is because HyperFlex uses scripts to configure everything for you. Management, vMotion, Storage VLANs, etc. It also spins up it’s own Virtual Distributed Switch and places it on a vCenter of your choice. Pretty cool.

In traditional networking you would need to uplink the HyperFlex to a couple of Nexus 7000’s and setup VLANs, then take those VLANs and Subnet settings and plug them into HyperFlex’s script. You can drag/drop that script into the UCS Manager and it will then populate all the required fields and then you can start the script running.

In ACI it’s a bit similar, however since ACI doesn’t so much care about VLANs, but rather End Point Groups (EPGs) you can pre-determine the VLAN names, and then feed it the appropriate Subnets. Just make sure those Subnets are attached to your EPG Bridge Domains and you are good to go! We also uplinked the HyperFlex into the ACI Leaf nodes in a Virtual Port Channel (vPC) configuration. So before we uploaded our script to the UCS Manager, we simply logged into the Fabric Interconnects and turned on vPC and Trunking, and then ensured all the VLANs were allowed on those interfaces and away we went. The script is good and makes everything soooo FAST and simple but there is a little prep work to do if you choose the vPC route.

Once the HyperFlex is up, you can go into vCenter and create your Storage, and link the vmknics to the proper EPGs in the case of ACI. As a note, ACI has several choices for a Virtual Distributed Switch that you can use. I typically go with the Application Virtual Switch (AVS) because it’s a VMWare product and we are using vSphere here. Recently however, VMWare announced that they won’t be supporting Cisco Products as before, so we might be using the Cisco vDS in the near future. Especially if we want to start using versions of VMWare greater than 6.2. That’s really ok though because now the vDS supports everything the AVS does, and actually for the Cisco Side it’s better.

As an aside I sometimes hear people talking about getting a 169.xx address on the OpFlex side for the VTEP. If this happens to use, be sure you:

  • Enabled Trunking on all FI’s
  • Enable CDP and disable LLDP throughout the Fabric. I mean EVERYWHERE, VMWare, UCS HFX, ACI or whatever you use.
  • Enable all VLANs required, especially your Infra VLAN for the Nics.
  • Ensure your Infra VLAN is setup in UCS/HFX as well.
  • For ACI Infra VLAN is automatically configured if using Routed Interfaces for BGP, otherwise, if you used say SVI interfaces, you will need to specifically allow that VLAN in your common VRF and allow THAT VLAN on UCS as well.

I don’t want to cover too much about the setup in this article, but rather convey my experience and hope that it may excite you to want to give it a shot. I can tell you that once I’ve used ACI and HyperFlex I never want to go back. I’ll cover ACI in another Article, but the benefits there are also amazing and with today’s Security Issues it is an absolute requirement if you want to quickly resolve patching hundreds or thousands of machines.

So you are probably wondering, what you need to get to use HyperFlex at this point.

Platform

This is the following hardware used in our setup. ACI is using a minimal configuration, and you can uplink to any other device as a simple access port upink providing it supports BPG, OSPF or some other typical protocol. ACI supports most of them so we used BPG as routed interfaces for the setup. For the HyperFlex we chose a minimum of 4. I want to be clear that you don’t need ACI to make this work. You can use your current network infra just fine.

  • 4 x Cisco UCS HX240c-M4SX (HyperFlex Requirement)
  • 2 x Cisco 6248UP 48p Fabric Interconnects (HyperFlex and/or UCS Requirement)
  • 3 x Cisco APIC M2’s (ACI)
  • 2 x Cisco 93180YC-EX Series Leaf Nodes (Tetration Supported) (ACI)
  • 2 x Cisco 9336PQ Spine Nodes (ACI)
  • 2 x Cisco Nexus 7000 (Uplink for BPG peering) (ACI)

Like I said, if you don’t have ACI yet, that’s ok. Just get the HyperFlex nodes and the FI’s if you don’t have them, then setup your network accordingly.

If this sounds interesting, drop me a line and I’ll point you in the right direction to get started!