Web Performance (WPI) released its latest investigative research on the leading virtualization software VMware ESX Server 3. Server virtualization is perhaps the most significant change in IT operations in the past decade, making it possible to manage groups of servers with a greater degree of reliability at a lower cost. For some IT operations managers, however, questions remain about how much overhead does virtualization add? How much slower does a single web application run in a virtualized server compared to a regular, non-virtualized server? The latest performance report from Web Performance takes a look at a typical PHP-based web application, and measures the performance both with and without virtualization in the premier virtualization product VMWare ESX.

The importance of putting real world benchmark numbers on an infrastructure is often overlooked", says Ben Miller, Chief Engineer at BlueLock, a company that provides IT infrastructure as a service. "Getting real data that can demonstrate the viability and scalability of an environment and application is crucial to proper planning for sizing and growth. In an infrastructure as a Service model meaningful application performance benchmarks will help a client to plan for their operating costs and predict with accuracy when they will need to scale on demand."

The test was set up with two identical software installations, one running directly on hardware, i.e. the traditional server configuration, and another running inside a virtual machine. Any server-based application would have sufficed; for the test the open source CRM package SugarCRM was used. The primary measure of performance used for the test is user capacity, which we define that as the number of simultaneous users that the system can support while meeting the specified performance criteria.

The results showed that the number of simultaneous users that could be handled by the virtualized server was 14% lower than the number of simultaneous users being handled by the traditional server configuration. "This test confirms again that the performance difference between real vs virtual machines is small enough to be made up for by an extra core", says Miller. HP's hardware and VMWare's Enterprise platform make adding this additional horsepower trivial and in most cases without downtime. The advantages in flexibility, business continuity, scalability, manageability, and the many other benefits of virtualization simply cannot be purchased for a traditional infrastructure at the small price of this virtualization tax."

Given the rate at which processor performance is increasing, this performance difference was neglible compared to the benefits of virtualization, and should put to rest any qualms about performance issues with virtualized systems. John Qualls, the CEO of BlueLock noted that "The test provides us an in depth understanding of the relationship between performance and the 'virtualization tax'. It gives us the confidence that something as simple as a $100 processor upgrade investment can mitigate this 'tax' while allowing our clients to enjoy efficient hosting and IT recovery thru a 100% virtual environment. This is important to BlueLock since virtualization is the enabler for us to deliver Infrastructure as a Service.