In the first part of this series on the top 5 virtualization problems for networking pros, Fast Packet blogger Josh Stephens explained the need for backup and recovery in a virtualized environment. But virtualization challenges don't stop there. In part 2, Josh explains the need to identify and battle virtual machine sprawl, which can crowd your network and make a simple technology way too complex.
Virtual machine (VM) sprawl sounds like something your dad does in his favorite recliner during the football game after the Thanksgiving meal, but this is an entirely different kind of sprawl and not even a bigger pair of pants will completely remedy the discomfort that it can cause.
You see, virtualization technology is easy, cheap and effective -- three features you rarely find together. But because it's so easy, it can get a bit out of control.
In the world of physical servers, there are built-in constraints to help combat sprawl. First, to deploy a server you have to find room in the budget, fill out a requisition form and then wait for delivery. Once the server shows up, it has to be re-imaged to meet your team's standards and then mounted in a rack somewhere. It will need power, cooling and a connection to the network. Ultimately, our data centers only have so much capacity for space, power, cooling and network access.
In the virtual world that all goes out the window. Need a new server? Just clone one of your existing servers that is similar to what you need and badda bing, you’re done. You don’t have to order, re-image or locate space in a rack. As a matter of fact, you probably don’t have to worry about cooling, power, network access, firewalls or any of the other infrastructure requirements because it’s all been done for the rest of the VMs hosted on that particular ESX server.
This is what leads to bloat (really just another word for sprawl), so here are three ways to identify and address VM sprawl.
1. Zombie VMs: No, these aren’t the cool flesh eating zombies that we love to watch Woody Harrelson shoot up or Bill Murray imitate. We’re talking about VMs that are sitting out there consuming resources but nobody really knows why. You kind of hate to delete them because you don’t know why they were created in the first place. I’ve often thought that “ghost VMs” would’ve been a better name, but “ghost” has an entirely different meaning in the IT world.
2. Orphaned VMDKs: No, these aren’t VMDKs whose parents were eaten by the zombie VMs mentioned above. These are Virtual Machine Disk Kernels -- effectively the files that still exist even though the VM they were attached to has been deleted. This is wasted space without doubt, but they can be a bit hard to track down. To me, they feel more like shadows than orphans, but I guess I’m not at the point in my career where I'm asked for advice on naming things like this yet.
3. Orc VMs: Yes, I made that up. What I’m talking about here are the VMs that started out good -- just as the Orcs started out as Elves -- but something has gone awry. These are the VMs that are being used by your business but seem to be reproducing like crazy and just like Orcs, they hate the sunlight. Try asking some of your application guys if they really “need” 100 VMs for the web server farm and you’ll see what I mean.
Controlling VM sprawl with management tools
We now have zombies, orphans and Orcs, and we need to get a tighter handle on things. That’s where today’s management tools come into play. All of the leading virtualization management applications help to identify and manage VM sprawl. Choose one that’s easy to use, maximizes your resources and provides a holistic view across compute, storage, network and applications elements of your virtual data center. Just don’t wait too long to get started; the longer you wait the larger the problem will become.
Read more Fast Packet bloggers
- Fast Packet blogger Greg Ferro debates: Cisco Cius vs. Ipad
- Seeking 10 GbE optimization? Fast Packet blogger Michael J. Martin says hardware alone won't help.
- Fast Packet blogger Ivan Pepelnjak declares both vCloud and VEPA fail in virtualization networking.
About the author: Josh Stephens is head geek and VP of technology at SolarWinds.