Best hypervisors for the desktop

Parallels Desktop for the Mac and VMware Fusion for Windows

Desktop hypervisors offer companies a way to run multiple operating systems, and the applications tied to them, on a single client machine.

Common use cases would be support staff, help desk and software development in a multi-operating system environment, or users with a specific business requirement to run a foreign operating system application.

We tested desktop hypervisors from Parallels, VMware, Oracle/Sun and Microsoft on both Windows and Mac hosts. Our guest operating systems were Ubuntu Desktop 8.10 and Windows XP SP3.

Parallels won our overall performance tests on both Mac and Windows platforms. And Parallels Desktop is our Clear Choice Test winner for Mac desktop virtualization. However, Parallels Workstation for Windows is a less evolved product, particularly when it comes to application sharing and guest virtual machine (VM) integration.

VMware's products are strong on both platforms. VMware Fusion (Mac) came in second by a whisker to Parallels, and VMware Workstation wins our Clear Choice Test for best desktop hypervisor for Windows machines.

Oracle/Sun's open source VirtualBox had decent performance numbers, but is more limited in its application execution capabilities between host and guest VMs, and is more difficult to use overall.

Microsoft's VirtualPC for Mac is no longer supported so we didn't test it. Virtual PC is destined to be the hypervisor that will enable Windows 7 to support Windows XP guests, but Microsoft doesn't seem interested in supporting non-Microsoft guests on Windows machines.


We installed a desktop virtualization application on the host operating system, and then installed a legally licensed copy of either XP or Ubuntu Desktop. For convenience, you can use a disk image file (we did) or a vendor's CD/DVD of the operating system.

Some of the desktop hypervisor products "recognized" that the guest to be installed was XP or Ubuntu and made automatic adjustments to accommodate various features between host and guest VM. This allowed us to make rapid default choices, such as how the guest operating system's disk storage would be made (and/or emulated), and how the native and guest operating systems might interact.

These interactions might be something as simple as shared folders among native and guest operating systems, or as sophisticated as being able to present applications as though they were "native" (while actually resident on the 'other' operating system).

To match our business cases, we tested each hypervisor with Microsoft Windows XP (32-bit) because of its popularity with Mac users. We also tested with Ubuntu Linux 8.10 (32-bit and 64-bit) to check support for Linux in general and to also have a 64-bit test to add to the mix.

We tested each hypervisor to compare features such as mirrored host/guest folders, application sharing, seamless presentation of guest/host applications and importing VMs from other hypervisors.

We discovered that the installation of each of the hypervisors was simple, and options for subsequent use plentiful.

VMware Fusion and Parallels Desktop include some extra applications to work with Windows XP or Vista VMs, and each include optional antivirus software (McAfee VirusScan Plus for Fusion and Parallels Internet Security powered Kaspersky for Parallels).

The security tools provided by Parallels also include a firewall and spam filter. Parallels also includes licenses for Acronis True Image Home and Disk Director Suite to help with backups and Windows optimizations, but are a separate download. These tools and extras weren't tested.

Gauging performance

We tested performance using two Java-based benchmarks. SPEC's SPECjbb2005 is a business-based benchmark that emulates a warehouse tracking application, uses no network I/O and little disk I/O. Instead, it's a better gauge of memory allocation, task forking and CPU muscle that's not affected by external (to the system under test) I/O.

The second benchmark used was SPECjvm2008, which touches numerous elements of performance and can be considered a better way of comparing desktop rather than server functionality through a series of discrete benchmarks that are run consecutively. We used three of the suite’s benchmarks: crypto, mpegaudio and xml.

The results showed that when Ubuntu Linux was the guest operating system, VMware Fusion beat Parallels. But Parallels returned the favor when XP was the guest. The results were close, and VirtualBox also did well.

Overall, Parallels Desktop was the performance winner, followed closely by VMware Fusion and VirtualBox. These results are based on the default settings, and it's possible to increase performance in any number of ways, which we didn't do.


Mac users were the first to express a strong need to use Windows applications, and this early demand is reflected in the fact that the Mac-based desktop hypervisors are generally more mature than those for Microsoft's XP.

VMware is the most consistent and feature-filled product on both Mac and Windows host platforms. Parallels was tops on Macs and an excellent performer. We can't recommend VirtualBox to the average user, and we can't recommend VirtualPC unless your installations will be homogeneously Microsoft Windows-based.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Must read: 10 new UI features coming to Windows 10