In this special Fast Packet series, blogger Josh Stephens outlines the top five virtualization problems. In the first three parts of this series, Stephens outlined issues with virtualization backup and recovery, VM sprawl and virtualization capacity planning. No. 4 in the list of top five virtualization problems? VM stall.
Virtualization problem No.4: Establishing trust and avoiding VM stall.
Now that you’re hip deep into virtualization and 30% or so of your servers are virtualized, it’s time to start thinking about moving to the next level and placing core applications onto VMs. The problem is, administrators of those applications and databases may drag their feet if they don't trust your virtual environment. Welcome to VM stall.
Most companies’ virtualization efforts follow three phases. During the first phase, new application servers are deployed onto virtual machines. Generally these are the application servers that are “easy” to virtualize or that pose little to no risk of being moved to virtual machines (VMs). Over time, because virtual servers are so easy to deploy, the number of VMs increases significantly and you end up at the 30% number I mentioned above.
In phase two, however, key infrastructure servers are migrated onto virtualization technology. This usually involves corporate email servers and database servers. At this point, VMs are the default platform for application servers, and physical servers are the rare exception. Phase three is when you start getting all cloudy, and we’ll talk about that in my next and last top five virtualization problems post.
Moving from phase one to phase two is largely a matter of establishing trust within your organization. Think about it: If you were the administrator for your company’s billing database, and some yahoo from the infrastructure team wanted to move your SQL Server from the fast, reliable physical system it’s been on for two years to a virtual server leveraging a shared SAN, wouldn’t you want some assurances that performance and reliability wouldn't suffer?
If you start making changes and application performance suffers, you’re going to be in the hot seat. And things can go south in a hurry when you’re talking about shared compute resources. If trust isn’t established, VM stall occurs.
The key to establishing and maintaining trust both with your peers and for yourself, and to avoiding VM stall, is solid performance management and monitoring. Managing virtual infrastructure and virtualized application servers require specialized techniques and tools. From a process standpoint, consider the following:
- Establish baseline application performance before migrating physical application servers onto VMs.
- Conduct weekly performance assessments after migration to ensure that any performance degradations are understood.
- Stay on top of bloat, orphaned VM Disks (VMDK) and zombie VMs.
- Perform real-time capacity optimization where needed to ensure that you’re getting the most out of your infrastructure.
- Be sure that you can see the whole picture—many times application performance issues in virtual environments are actually caused by SAN I/O or network performance.
If the processes above scare you, then you’re probably using the wrong management tools. Specialized technologies like virtualization require specialized monitoring tools and techniques, so look for tools that make it easy to perform processes like the ones listed above.
Solid management is the key to establishing trust. There’s a lot of upside to doing this right and continuing to advance your adoption of virtualization technology, but there are also risks to doing it wrong, so be sure to make smart management a part of your deployment and migration strategies.
About the author: Josh Stephens is head geek and VP of technology at SolarWinds.