For the past few years, the big boys (HP, Cisco, Dell) have been trying to convince us that converged infrastructure is the way to go. Buy the stack of compute, storage, and network from us and off you go. One throat to choke, the best experience because all the components are engineered to work optimally together.

But is converged infrastructure already on the way out and now hyper-converged infrastructure the way to go with virtualized workloads.

My first experience with hyper-converged was with Nutanix. To be honest I didn’t get and didn’t trust it. I needed to get my hands dirty in order to understand the new approach to virtualized workloads that Nutanix offers.

It finally clicked and then eventually I find there are others out there that compete in this space (Simplivity, Scale Computing, Pivot3). Great! I am all for competition and choice.

When it comes to hyper-converged solutions in general, some of my concerns are:

  • Will the solution start cracking apart at a certain scale? The file system starts to crack at certain node count.
  • And since I am a desktop virtualization guy, does the solution support GPU’s?
  • Do you support only a single virtualization platform? Will there be additional options in the future?

From an administrator point of view, hyper-converged solutions are a great solution because of the brick by brick ability to scale out approach.

What’s been your experience with hyper-converged solutions? Are you using it for a certain use-case?

After installing Cisco UCS and performing the initial setup of the UCS Fabric Interconnects; there is a lot of work that needs to be completed prior to configuring the blade Service Profiles.

This can be done a multitude of ways, manually through UCSM, scripted through SSH or now through PowerShell using Cisco PowerTool. This allows us to use variables at the top of the script; these variables allow us to make the script portable for various customer deployments.

There are a couple features that are not working yet:

  1. Setting the Global Power Allocation Policy – due to the enormous lack of detail in the get-help sections of Cisco’s PowerTool, I have been unable to find which function controls the Global Power Allocation Policy.  I believe it to be the set-ucspowergroup cmd-let but I have been unable to make it work correctly.
  2. Adding custom roles with customer permissions – the add-ucsuserrole cmd-let is not very well documented either, like others; I cannot seem to figure out how to get this cmd-let to create a new role with custom permissions.

When building out your UCS deployment, you create a series of Fibre Channel (FC) uplinks. Then you have to go back and add those FC uplinks to the VSAN’s that you created. This can be tedious if you have a lot of FC uplinks as you would need to modify each one. With Cisco’s Powertool, you can do this quite quickly with no mouse movement :)

As you can see:

are variable used so that you only need to modify the VSAN ID once. Otherwise, you would need to specify on each and every line.

This took me a while to figure out as Cisco’s get-help files (man pages for Linux folks) leave a lot to be desired.

if you run:

You get:

Which would lead you to believe that this command would work just fine:

But it does not, you have to pipeline through the get-ucsvsan output first.

Creating QOS policies within UCS Manager can be somewhat cumbersome. In UCS Manager, on the Lan tab; you would right-click UCS policies and click create… one by one you would go through this process to define you UCS policies, if you have a lot of policies or a lot of UCS builds to do this can become quite monotonous.

Creating Cisco UCS QOS Policies

Creating Cisco UCS QOS Policies

Using Cisco PowerTool, this can be done quite quickly using the Add-UCSQosPolicy and Add-UCSvNicEgressPolicy command-lets. For example:

Would create the Best Effort, Bronze, Gold, Platinum, and Silver QOS Policies.