Server virtualization metrics and virtualization benchmarks are critical to analyzing the financial and operational performance benefits from virtualization. This is especially true when you want to know if you are achieving the full ROI that virtualization offers for improved staffing efficiencies and speed of execution.
Unfortunately, there is a lot of confusion on which server virtualization metrics to use, how to calculate server virtualization metrics and how to locate the appropriate virtualization benchmarks. And the problem is just as bad if you are looking for desktop virtualization metrics and virtualization benchmarks.
Since I just went this this exercise with an organization I thought I would share it here. Like all IT metrics and benchmarking strategies I always advocate looking at things holistically using an IT balanced scorecard approach. This is especially the case when you are trying to validate the post-implementation ROI of a server virtualization project, VDI implementation or any other desktop virtualization solution using specific virtualization metrics.
So here are what I think are the most useful virtualization metrics, how to calculate them and a bit about their meaning and potential use in your IT balanced scorecard.
Percent of Servers Used for Virtualization
Description: The percentage of the physical servers that are being used for virtualization.
Formula: Servers used for Virtualization = Servers with a Hypervisor ÷ Total Servers x 100
Use: Perhaps the most frequently mentioned of the virtualization metrics is this widely used industry metric that tracks the percentage of servers being shipped that will be used for virtualization. This makes it easy to track your performance against the industry as a whole. The real value of this metric comes when it is used in conjunction with physical server utilization data and CapEx spending. A low percentage likely means the opportunity to lower your CapEx spending still exists especially if you have low utilization levels making it a core indicator among financial performance metrics.
IDC estimates the percentage of all servers purchased that will be used for virtualization at around 30%. But several studies that exclude Google (1 million servers), Facebook (100K+ servers) and some other big shops puts the actual percentage closer to 70%.
Virtualization Penetration Rate
Description: This is often simply referred to as Virtualization Rate or what VMware calls Percent x86 Servers Virtualized. It is the percentage of all servers that are virtualized and is calculated as a percentage of all servers physical and virtual.
Formula: Penetration Rate = Virtual Servers ÷ (Virtual Servers + Physical Servers) x 100
Use: Among the virtualization metrics you can see why it is a better indicator of how well you are leveraging virtualization technology to improve efficiency. Some argue the physical servers running a hypervisor should be excluded from this calculation. But, even VMware does not do that in their performance metrics model because every physical server represents an actual cost (CapEx and OpEx) regardless of its role.
So if you don’t count all physical servers, several performance metrics will be skewed including several important financial performance metrics. One industry group’s most recent survey puts the server virtualization penetration rate in the US at 43.4%.
Percent of Virtual Desktops
Description: The percentage of the total number of client computing devices (desktops or laptops) that are virtualized which reflects the penetration of desktop virtualization.
Formula: ‘Desktop’ Virtualization Rate = Virtual Desktops ÷ Total Desktops x 100
Use: The percentage of desktop virtualization is not only a measure of virtualization penetration it may well be a strong indicator of the IT organization’s ability to respond quickly to new demands. Like other virtualization metrics, desktop virtualization rate represents an IT capability to quickly deploy changes to the users including changes to multiple device types.
VM to Admin Support Ratio
Description: This metric compares the total number of virtual machines to the number of FTE’s required to support them, which may be a fractional FTE.
Formula: Number of VM’s vs Number of VM Admin FTE
Use: This may be one of the most common virtualization benchmarks people look for. As an organization moves more systems from physical servers to virtual machines, they should expect to see the number of FTE’s required to support the virtual machines go down in absolute terms.
The benchmarks here are difficult to sort through because some performance metrics sources include large hosting providers where the ratios can be 1000:1 or more. There is also the typical scaling effect where larger corporations have more virtual machines and therefore can have ratios of 250-400:1 whereas smaller organization may only have 100-150:1. This is why it is important to allow for fractional FTE’s in this ratio in smaller shops where the admin likely splits their time between other duties.
Percent of VM’s Covered by DR Protection
Description: This metric is the percentage of virtual machines that are supported by the disaster recovery plan.
Formula: VM DR Percent = VM’s Covered by DR ÷ Total VM’s x 100
Use: This metric is used as an indicator of what types of uses are being run on virtual servers. VMware views a low percentage as an indicator the virtual servers workloads are not Tier 1 and estimates a potential savings of $500 per hour in avoided downtime in their virtualization ROI for any Tier 1 outage.
A clear benchmark is not available. Using an internal benchmark as baseline may be just as effective.
Days to Provision a VM
Description: This metric represents how many days it takes to provision a virtual machine including the entire request and procurement cycle.
Formula: Average Number of Days Required
Use: This metric represents a service performance metric which can be improved by automation and process maturity to create a more repeatable process with consistent results. In simple terms this is a speed of execution measurement.
A clear benchmark is not available. VMware does indicate in their assessment model that more mature organizations can provision virtual servers a couple of days faster than those with less process maturity.