Manage Learn to apply best practices and optimize your operations.

Best practices for physical environment tests before setting up virtualization

Performing physical environment tests in a thorough manner, determining what you should and shouldn’t virtualize, will smooth the process of setting up virtualization on customers’ servers.

Before starting any virtualization project, it’s important that you help your customers thoroughly understand their current physical environment.

Introducing virtualization into their environment will have a ripple effect on all parts of their infrastructure, including standard procedures such as monitoring, backups, patching and administration. As a result, you need to assess all parts of their infrastructure -- not just the servers you plan to virtualize that will let you uncover any potential problems or hurdles that may affect your project. The old woodworking rule “Measure twice, cut once” also applies with physical environment tests. You can help save your customers from making costly mistakes by making sure you help them get accurate measurements before you begin.

Physical environment testing and assessment doesn’t just mean gathering performance data; there are other areas that you need to consider as well. A good pre-virtualization assessment will cover the following:

Documenting the current physical environment of the data center
Thoroughly documenting a customer’s current physical environment before attempting to migrate it to virtual servers ensures that the proper-sized server hardware and correct number of virtualization licenses are purchased. It’s a good idea to thoroughly inventory all of their current physical servers so you know exactly what is there. Also, you should find out what they want to do with the old physical hardware after it has been virtualized. From there you can help them decide if they want to reuse newer physical hardware as ESX hosts, leave it as is or discard it. Good documentation will help prevent any surprises from occurring during their project, which can cost your customers additional time and money.

Know your resource requirements
You should gather performance statistics from the servers or desktops that you plan on virtualizing. This ensures that your virtual environment is sized properly to handle the physical servers that will become virtual machines (VMs). VARs don’t need to pull every performance statistic, but should know the basics, including average and peak usage in each of the four resources areas, CPU, memory, disk and network.

Solution providers should collect performance data over time to catch any trends or spikes in resource usage that may occur during specific time intervals. You should gather data for a week at minimum, and preferably a month. The combined results of your performance statistics will help determine your host sizing requirements.

Enterprise monitoring systems that can centrally collect this data will make performance data gathering and reporting much easier. There are also a number of products specifically designed for physical environment tests and assessment. VMware, for example, has a Capacity Planner product that is available for free to solution providers and there are also a number of more thorough third party applications available, such as PlateSpin Recon.

Decide what you want to put in a virtual environment
With the capabilities and scalability of today’s hypervisors, almost any workload can be run in a virtual environment. But it’s rare to find companies that have their physical servers 100% virtualized, so it’s up to you to help determine what should be migrated. Keep in mind that virtualization is about more than just consolidation and there are many other features and benefits that it provides as well.

Many companies start off by virtualizing their low-hanging fruit, which is typically print, file and utility servers, before moving on to servers that run more complex applications such as Exchange Server. This approach gets them experience with virtualization on low-impact servers before they start virtualizing their more critical systems.

Some common reasons for not virtualizing physical servers include high-resource utilization, licensing issues and application support issues. But most of these can be mitigated because hypervisors such as vSphere are scalable with low overhead from the virtualization layer. While you may not get high consolidation ratios when virtualizing high-resource workloads, you still get the benefits of virtualization. Don’t let fear of licensing issues deter you, as most vendors have changed their licensing models to be virtualization-friendly. The same holds true for application support as most vendors now support running their applications in a virtual environment.

Many customers still run legacy applications on old hardware and using very old operating systems (OSes). These types of applications are usually ticking time bombs because they are no longer supported and a hardware failure could permanently seal their fate. Virtualization is ideal for these types of applications as they can run forever as VMs, regardless of the underlying hardware. Virtualization encapsulates a guest OS that uses virtual hardware that can be run on any physical hardware without affecting the application and guest OS.

Decide what you don’t want to virtualize
As mentioned above, there are some legitimate reasons for not virtualizing servers, one being support issues. Some application vendors will make a best effort to support applications in a virtual environment, but may ask that you reproduce the problem in a physical environment. That’s why it’s a good idea to keep at least one physical server around for applications that typically run on many servers. If you, for example, have eight domain controllers in your environment, you may consider virtualizing six of them and leaving two as physical servers. The same goes with database servers. If you leave one or two Oracle or SQL database servers as physical servers, it gives you the flexibility to move a database hosted on a VM to a physical server if the vendor requests it.

The physical environment also helps keep vital infrastructure services running in the event that your environment suffers a major failure that may cause you to lose most of your VMs. For this reason you may want to copy at least one Domain Name System (DNS)/ Dynamic Host Configuration Protocol to a physical server because almost everything in the data center relies on DNS to work properly. The same holds true for other critical services like domain controllers and authentication servers. Having at least one physical server for these services ensures that they stay up if your virtual environment has problems. Another note to keep in mind is that since physical servers don’t have the same dependencies as virtual servers, they can be brought up much quicker in the event of a total outage.

Other considerations
Virtualization isn’t going to magically fix any problems that may already exist in your customer’s physical environment; it may even magnify them and make them worse. If an application runs poorly in a physical environment, it’s probably going to run poorly in a virtual environment. It’s best to resolve any application issues that may exist before moving it to a virtual environment.

Your customers need to change their way of thinking with a physical environment because virtualization is all about making the most efficient use of resources and not wasting them. Just because a physical server may have four CPUs and eight GB of memory doesn’t mean it needs that same hardware in a virtual environment. While wasted resources maybe acceptable in a physical environment, it defeats the purpose of virtualization, where you should right-size your VMs and give them only the required resources. Start on the low end when assigning resources to VMs because virtual hardware makes it easy to change at any time if you need to add more resources to a VM.

In addition to hardware, assessing a customer’s environment should include software applications. You should do a complete inventory of applications running on the servers that they plan to virtualize so they can ensure there will be no support or licensing issues when running them on virtual servers. There may also be special licensing considerations or configuration changes that need to be made to a virtualized application.

Put together a spreadsheet of all of your customer’s applications and check with each vendor to find out their virtualization support policy and see how they license their product in a virtual environment. You can often find this information out on their Web site or in their knowledge base. Include columns for their support level, licensing model and URLs to their policies. You should also meet with the application owners to discuss the results and make sure they understand the vendor’s virtualization policy.

Not performing a physical environment test and assessment prior to server virtualization can lead to cost overruns, time delays and potential project failures. Planning is always crucial and because virtual environments are complicated, there is no room for making assumptions or educated guesses when architecting them. A proper assessment will ensure your customer’s virtualization project goes smoothly without any hidden surprises.

About the expert
Eric Siebert is a 25-year IT veteran whose primary focus is VMware virtualization and Windows server administration. He is one of the 300 vExperts named by VMware Inc. for 2009. He is the author of the book
VI3 Implementation and Administration and a frequent TechTarget contributor. In addition, he maintains, a VMware information site.

Dig Deeper on Server virtualization technology and services