Build a Better Action Plan for Managing Your Virtualized Capacity

Pin the Tail on the Donkey is a lousy game at your kid’s birthday party and it’s a lousy way for your organization to do its virtual capacity management.

Doesn’t it feel like Pin the Tail on the Donkey when you’re doing resource planning for your virtual environment? John Maxwell posted a couple of weeks ago about some action-oriented ideas you could put into play that would allow you to better manage your virtual capacity  and model workloads. However, that is easier said than done if you don’t have the proper tools in play. 

What to look for in virtual capacity management

A recent Forrester report, The Case for Virtual Capacity Management, shows that only one in four virtualization administrators uses proper capacity planning tools and modeling. For all the rest of us, here are the most important areas in managing resources and virtual capacity:

·        Current capacity

Current capacity reveals two things: which clusters can accommodate more VMs, and how many more VMs can be created before reaching the limit of resources (CPU, memory or disk).

Take an environment with, say, 200 VMs running on 10 ESX hosts. Your slot size, or the resources allocated to a virtual machine, is the unit of measurement to use in planning. It can vary from cluster to cluster and shop to shop, since user needs and business needs differ. For that matter, it may vary from year to year within a shop as workload fluctuates.

Your sample slot size could be 4 CPUs, 2GB memory and 512 GB storage.

Administrators usually create a template for their slot size so they know how many more slots they have available; that is, how many VMs they can add before running out of virtual CPU and memory. Naturally, the lowest remaining resource is the gating factor in capacity management, so if you have enough memory for 20 more VMs but only enough CPU for 10 more, then 10 is all you can add.

·        Resource utilization

Resource utilization contrasts utilized resources against available capacity to show when you are likely to run out of resources.

Measure your utilization separately from your capacity. Your goal is to forecast how much CPU, memory, storage and I/O the virtual machines will require. If you’re executing perfectly, actual utilization is 100 percent of capacity, but nobody hits that all the time.

The virtualization engine borrows resources from any VMs that don’t need them and reallocates them to the ones that do. Suppose you’re running an ESX host with 20 VMs. When you create each one, you allocate resources to it (usually by using the template for your slot size). If you allocate four CPUs to the VM, but it needs only two of them, the virtualization engine borrows CPU resources for other, needy VMs.

·        Host Requirements

Host requirements measure the workload of VMs running on ESX hosts (servers). Understanding your host requirements lets you answer the question, “Do we have enough servers to support the current and future VMs in that environment?”

You can get so wrapped up in the virtual world that you forget to think about the physical world. Since your VMs depend on the physical servers underneath, plan at that level also.

  1. Look at the number of physical servers you have. How much physical CPU, memory, storage and I/O capacity do they have?
  2. Consider the workload you’re placing on them: How many VMs is each physical server running?
  3. Think about the workload over time. Will your physical resources suffice to host current VMs plus those to come? For how long?

As a result of that process, you’ll know whether you have enough servers to fulfill commitments now and in the future.

·        Server Refresh

Machines aren’t young forever. They get old and technology advances (solid state drives, more DIMMs, new generation of CPU). That applies to any server, whether it’s hosting virtual machines or not, and it’s part of capacity management.

Suppose you have a cluster of physical servers. Every few years, you need to replace them. Here, virtual capacity planning is a matter of deciding whether you want to replace them all at once or stagger replacement in batches over time. It’s a matter of virtual machine workload, the cost associated with buying new servers and the impact of retiring old ones.

New Forrester report: The Case for Virtual Capacity Management

Look at all four of those areas over time, then extrapolate. You may base your model for future capacity and utilization strictly on organic growth. Or you may foresee a bump in new users and more data being stored and numbers being crunched because of an upcoming acquisition. Or you can model on what-ifs, like what if Sales needs ten more VMs and Operations needs three fewer VMs? How do those business changes translate into changes in CPU/computer utilization?

We’ve posted The Case for Virtual Capacity Management so you can see how your virtualization environment stacks up against others. See the kinds of questions organizations like yours are asking about utilization and capacity management.

And dump Pin the Tail on the Donkey for your kid’s birthday. Nerf gun wars are much more fun!