Welcome to Virtualization and Beyond

Scaling and Sizing your Virtual Environment

By Chris Paap, Technical Product Manager, SolarWinds

It doesn't matter if you're an administrator over a modest virtual infrastructure of just a couple hosts with a few virtual machines (VMs) or if you have the gargantuan task of managing several hundred hosts with tens of thousands of virtual machines, properly scaling and sizing is essential for continued operational performance.

However, the key to doing this properly does not solely rely on predictive analysis and having concrete numbers of what workloads will be added at a future point in time. In fact, as most admins would attest, we usually lack the correct data to make accurate growth predictions. So, aside from a magic crystal ball to make accurate predictions, what are the key elements to consider when the time comes to scale and size your virtual infrastructure?

Well, there is no one-size-fits-all way to scale and size an environment to accommodate for growth and maintain peak performance. Since every environment varies, the key is to identify your goals related to what you're trying to accomplish.

For example, high availability (HA) and fault tolerance are requirements of most production environments, but what differs is the SLA time that is acceptable to recover from an outage. How you provide HA and fault tolerance should be a factor in determining how you scale.

In the past, most admins would scale up with large hosts and pack these with as many VMs as possible so they would have a few very large hosts with a high density of VMs. However, when a high density server suddenly went down, all those VMs then had to start back up on another host in the cluster. If you have this model, but also have an SLA requirement with very little downtime during an outage (such as a hardware failure on a host), you must ensure you have the underlying network, server and software infrastructure to accommodate fast recovery.

For this reason, many admins have started to scale out verses up to not only reduce the risk of a single host going down and interrupting a high number of VMs, but also because of the adoption of converged infrastructure.

The type of technology being used in your virtual environment plays a role in scaling out as well. Converged infrastructure combines network, server and storage into a single unit, and hyper-convergence takes this a step further by taking a software-centric infrastructure that integrates all these components into commodity hardware that is grown in a scaled out fashion. Cloud also takes this a step further since hardware is no longer the commodity, but instead CPU cycles, storage IO and network IO are.

As obvious at it might seem, planning is a key component to scaling, yet architectural decisions are constantly made with a tactical milestone in mind as opposed to taking into account available technologies and your goals.

An often overlooked component of successfully scaling virtual infrastructure is monitoring. It's crucial you're validating and confirming that your performance is meeting or exceeding your expectations. And good monitoring should not just identify when performance is bad, but identify trends and optimization actions to take to ensure continued operational performance in order to get ahead of the curve. It's important to note that your monitoring infrastructure will need to scale alongside the virtual environment. Polling intervals, data retention times and the number of virtual objects being monitored all play a role in the sizing, capacity and performance requirements of your monitoring solution in order to keep up with growth.

How do you scale and size your virtual environment?

##

About the Author

With 14 years of IT systems engineering experience across multiple corporate environments, Chris Paap currently serves as a technical product manager for hybrid IT performance management software provider SolarWinds, where he focuses specifically on the award-winning SolarWinds Virtualization Manager. In this role, he is responsible for defining the product roadmap and identifying key new features to solve IT problems.