Virtual Cloud Strategy

Dig Deep Into Your Hypervisors

Don't take your hypervisors for granted. If you know them at a surface level, you don't know them at all.

• By Greg Shields, Don Jones
• 08/28/2014

For at least a couple of years now, virtualization has seemed to be getting less exciting. It's not like it was in the good old days, where the very fact that hypervisors existed was simply amazing, and building out high-performance infrastructure was really challenging. The big virtualization vendors these days are spending less time talking about hypervisor features and more time talking about management tools (that is, when they're not all using the word "cloud" ad nauseam). The hypervisor, it seems, has become about as interesting as an unmanaged 10baseT hub.

But we think now is exactly the time for the hypervisor to be more interesting than it has ever been. That's because, in large part, the big headline features for hypervisors are all here. We're not likely to see massive low-level changes along the lines of live migration, memory overcommit and so on. The big things exist; what we'll get now are a continual stream of refinements and tweaks, combined with lots and lots of management tools.

So if the big features are already here, what's there to get excited about? With the big features rolled out and fairly well understood, now's the time for you to become incredibly smart about how the hypervisor actually works -- deep down, under the hood, at a primal, nitty-gritty level. How, exactly, does memory overcommit work? How does it affect applications running inside the VM? How does a given application respond to a Windows OS low-memory condition, and will it affect that application's overall performance? How does live migration work? What exactly has to happen under the hood for a VM to magically "move" to another host? What specific host hardware features can help or hinder that condition? What would you monitor to troubleshoot slow migration performance?

If you started early on in virtualization, you probably knew a lot of under-the-hood details about how the hypervisor worked, because in large part you were exposed to those details on a daily basis. As the feature set became bigger, it probably became easier to ignore some of the details -- especially as the administration and monitoring tools got better at insulating us from those details. Today, it's incredibly easy to spin up VMs, migrate them, configure them and do everything without ever seeing what's happening behind the scenes. You just know there's an admin someplace who doesn't know the difference between a VMX file and a BMX bike. That's dangerous.

The big virtualization vendors tend to focus less on the low-level details, even at their own educational conferences. That's because you've already bought their hypervisor, in most cases; what they're trying to get you excited about is the next purchase, which is probably tooling that sits on top of that hypervisor. Not that tooling isn't important, but if you want to remain a high-tier, well-paid virtualization expert, you're going to need to keep your head in the low-level details, as well. More important, if you're a decision maker in charge of a virtual infrastructure, you need someone on your staff who has the low-level knowledge needed to plan, architect, troubleshoot, and monitor your infrastructure. Tools are great, but at the end of the day, it's that super-deep experience that makes all the difference.

Point being: as a strategic measure, don't get so caught up in the high-level tools that you lose sight of what the low-level stuff is doing. Now's the time to double down and become a real expert about the guts of your hypervisors.