You've heard of "rightsizing," right? When it refers to an organization, it heralds a stressful time waiting for that pink slip. When it refers to a virtual infrastructure, it also heralds a stressful time, though without the pink slip.

 

Let's say you're in charge of your organization's virtual infrastructure. Some people complain that their VMs are slow - the CPU or the memory is consistently pegged when they're using the VM - and they can't work. Other people have no issues with their VMs; in fact, the VMs they used are lightning fast, and their usage barely makes a difference in the CPU. Perhaps when you open your VM manager, you notice that you have significantly more VMs than people.

 

The Problems

 

While there are any number of reasons for these issues, though no user would call "lightning fast" VMs an issue, for the purposes of this post, let's us reduce these issues to underallocated VMs, overallocated VMs, and VM sprawl.

 

Underallocation

A virtual machine does not have enough resources allocated for its current usage and runs into performance problems. Either the usage profile has changed or the VM never had the correct amount of resources. In general when performance is slow, the memory or CPU is the culprit.

 

Overallocation

When a VM has more resources than it uses, the VM is overallocated. Many people do not consider overallocation to be a problem, and some VM administrators even think they're doing someone a favor by setting up a super beefy VM. While it's not a problem for the user, it will eventually be a problem for the other people using VMs on the same host since resources are reserved that will not be used. A pie analogy works well here. There is a finite amount of pie, and you have taken two slices when you only want to eat one.

 

Sprawl

VM sprawl occurs when more VMs are deployed than are used or needed, thus reducing the total amount of resources available for other VMs. This is unique to virtual environments, as VMs are considered "cheap" to create and few remember to delete VMs that they no longer need.

 

The Solution

 

These problems can be solved with a judicious application of "rightsizing". Merriam Webster defines rightsize as "to undergo a reduction to an optimal size". In virtual infrastructures this means reallocating resources so that each VM has exactly what it needs and reducing the amount of resource draw by VMs that are not being used.

 

For underallocated VMs, this means adding resources or, possibly, moving the VM to a different host.

 

For overallocated VMs, you reduce the amount of resources a VM uses to free those resources for use by other VMs.

 

For VM sprawl, you identify the unused VMs and remove them, including any leftover files.

 

Going through each host, cluster, or VM manually to check the resources, or conducting user surveys on the performance of their VMs is time consuming and not particularly cost effective. Instead, use one of the many VM management tools to help identify these types of problematic VMs.

 

Coming to VMware monitoring, you can see this Tech Tip from SolarWinds to learn about how SolarWinds Virtualization Manager, the ideal VMware monitoring software that can help you rightsize your environment. VMware performance monitoring, simplified!