Although Fusion is still in beta, some of its features compare favorably to Parallels.
For the past few weeks, I've been in a virtual nirvana of sorts, working with both the latest Parallels release and the second VMware Fusion beta.
Both products provide an interesting contrast in the things they focus on. For right now, Parallels is all about Windows. It supports Linux, but you lose Coherence (the 'rootless' feature, where your VM applications appear to coexist with your Mac OS X applications), and a number of other conveniences. VMware has more mature support for Linux on tap for Fusion, which is no surprise VMware has a more mature product line. However, Fusion is still clearly in beta. For example, I still cannot get the VMware tools for Linux to install in my Ubuntu VM.
Even in its beta status, VMware Fusion for Mac OS X shows the advantages of having richer heritage. For example, even though it's not perfect, there is accelerated 3-D support in Fusion. This is important not just for the obvious use of game playing in a VM, but for running 3-D and other graphics - intensive applications that aren't on the Mac yet, like 3D Studio Max, Autocad, Expressions, Acrobat 3D and Designer, etc. Yes, that's right, 3D isn't just for games anymore.
Another benefit of VMware is the ability to assign CPUs to a VM. Now on a dual-core system, like my MacBook Pro, this isn't such a big deal. But on a Mac Pro, which is running four cores, this can come in rather handy, as you can now assign cores to VMs as needed. If you want to run multiple VMs, (and who doesn't), this feature doesn't suck.
Both implementations have a "Physical to Virtual" system, Transporter for Parallels, and VMware Converter for Vmware. Both do essentially the same thing, help you convert a physical machine to a VM. Both can also convert other VM's to their own native format.
One advantage here for Parallels is that it can convert VMware images, whereas VMware converter doesn't (as of yet) deal with Parallels images. Neither handle Linux or Unix directly. The advantage VMware has is in scale. You can use converter either locally on a Windows desktop, or you can go to VMware's Virtual Appliance Marketplace, and download preconfigured VMs. Need an appliance, (VM) preconfigured with Zimbra's messaging environment? Done. Need an Oracle 10g appliance? Done. That's part of what you get when you're the little brother to something like VMware ESX, VMware's enterprise offering.
Fusion and Coherence
Running Fusion is, for now, a mix. On the user feature count, I'd have to say that Parallels is still out in front. Aside from the things in Fusion that are still in the "Not yet implemented" state, Parallels is just better designed to fit in with your Mac OS X desktop.
The big reason here is Coherence. Being able to just click on a Windows application in the Mac OS X Dock, and have it come up is A Good Thing. However, Fusion wins for me in how it treats your OS X system. First, unlike Parallels, it doesn't install various kernel extensions in /System/Library/Extensions. So, if Fusion decides to fall down and go boom, it's harder for it to take your Mac OS X host operating system with it.
I know that on a "feel" basis, my Mac OS X host OS "feels" more responsive when I'm running Fusion than when I'm running Parallels. Parallels, at least on a first-generation MacBook Pro with 2GB of RAM, puts a whack on your system. Regular pauses where nothing's quite working, switching between Mac applications can take a long time, etc. In comparison, Fusion seems to abuse my Mac OS X host environment far less. Of course I don't get Coherence, and there's this odd bug wherein if you change primary IP addresses, the vmnet-bridge process starts complaining bitterly to system.log at a fast clip. That however, is to me, a beta bug, so I'm not to worried.
While I don't know for sure every aspect of who Fusion is targeted at, for now, it appears to be more of a product for those who only need to run a VM or multiple VMs occasionally, while Parallels seems to be more for those who need to have their primary VM running constantly. While I'm positive that Parallels doesn't abuse a Mac Pro nearly as much as my laptop, I'd hesitate, based on my experience, to run multiple VMs with a lot of Mac OS X applications running, especially Rosetta applications.
While it's not soup yet, the latest beta of Fusion is showing a lot of promise. It runs well, and doesn't abuse my system any more than it has to. That's always good.
However, there's the elephant in the room. Virtualizing Mac OS X and Mac OS X Server. There is no way to doubt this is both desired and useful, however there are issues. First, unlike some, I don't have a great need to run Mac OS X on non-Apple hardware. For my world, there's no cost savings. By the time we get done spec'ing out a Dell or HP 1U rackmount server, the price advantage over an Intel Xserve is negligible to non-existent. Others will disagree, but I don't see Apple ever changing their EULA or their OS to allow it to run on non-Apple hardware. Apple is primarily a hardware company. This would be a silly move for them.
But there is real, immediate value in being able to run, on Apple hardware, an ESX implementation, so I could have multiple Mac OS X or Mac OS X Server VMs running, with or without other OS VMs. Being able to up my hardware utilization to that degree would be a huge win for me, or any other Mac administrator.
Sure, I'd have to replace my PPC Xserves, but so what? That's an instant ROI win. When you look at the overall feature set for ESX, such as redundancy, high-availability, remote live backup ESX servers, and all the rest, ESX running on Apple hardware would do wonders to help Apple in some areas where they're weak, and may be weak for some time to come. Of the two companies, I think VMware will be able to do this first. They have the maturity and experience at the technical and business level to work with Apple's understandable hesitancy in this area. And I don't see them having a particular problem with making sure Mac OS X and Mac OS X Server VMs only run on Apple hardware. It may not be their desired goal, but it's still a business win for them.
In conclusion, while Fusion is still clearly a beta product, it's much improved from the initial release, and shows great promise.