Ever consider how are your VMs are being treated?
When VMware released vMotion it was revolutionary and amazing. It was like magic! Then, in 2006 with VI3 we saw the rise of DRS. Wait…VMs can move on their own, without humans? It was the singularity! VMs moving around to find the resources they need, all by themselves.
Enter 2011…
Over the last nine months, I have spoken at seven VMware User Group (VMUG) events. I always ask the same three questions and almost always receive the same response. How many of you use vMotion? Every hand goes up. How many of you use DRS? About 10 hands usually stay up. How many of you let DRS move around VMs without any Human interaction? Only one or two people remain with their hands up. At VMUG Seattle, five pioneering engineers were left with their hands raised high.
So why this disconnect? Isn’t that what virtualization is all about? Well, there are a few things to consider.
First, your data center is carved out into islands by routing protocols. If you’re wondering whether this is a problem, allow me to put things in perspective.
Let’s just say you buy yourself some nice, shiny new VMware. You’re excited to try to do your first vMotion until…wait! What’s this?! Each of your racks is a separate “/24″ Network? You discover your VMs are only capable of chasing their own tails around and around in their own rack? Oh the humanity!!
Photo courtesy of: Cory Doctorow
Now, on to the second issue: Keeping VM performance up and IT issue tickets and phone calls down.
Suppose you were to move a VM from one rack to another, or even to another row. What if the performance of that VM was .002% less than it was at the original location? Just picture it now…15,000 phone calls with angry folks yelling, “WHAT DID YOU DO!” – and while you watch your inbox fill up!
And what about vDS? Oh sure, vDS can do a great job of making sure QoS settings on hypervisor A are copied over to hypervisor B. What if it’s not a pure VMware environment? What if someone kicks off a back-up job? Or a DBA runs the monster of all SQL calls, resulting in the flooding of that same switch your VM is now underneath?
Pitchforks and torches! (Tickets and phone calls!)
Photo courtesy of: aneye4apicture
So instead of having “Free Range VMs” that happily graze your data center “cloud,” migrating from rack-to-rack to find the greenest server grass, you now have caged VMs in tiny 19″ wide 74″ long stalls. Scared and alone your VMs stand. Between gaping moats created by routing and a realization that it’s actually quite difficult to offer apples-to-apples performance when VMs roam on their own. As a result, vMotion is often relegated to a simple “move off, move back” mentality between servers in the same rack for hardware swaps during a maintenance window. DRS never even gets the chance to reach its full potential.
The dream of VMs freely roaming the green hills of your data center looking for a nice warm place to gnaw on memory has vanished.
So is there any hope?
Actually, there is! You have Ethernet fabrics, and AMPP!
Ethernet fabrics, using technologies like TRILL, take your silo’d data center and turn it into a vast green flat layer 2 network, free from the confines of router boundaries and free to roam. AMPP allows you to create a profile attached to a MAC, which can contain things like QoS info, VLAN tags and ACLs. Profiles follow your VMs around ensuring that regardless of where they roam they’re provided the same clean water, same green grass, and same warm sun they had before.
Realize the dream. Give your VMs the life they deserve. Happy VMs graze on Ethernet fabrics. Be Free VMs!
After all, no one wants “VM Veal.”
