Felines, Fluids, And Laptops: The Virtualization Followup

With my new Windows Vista-powered laptop currently acting as a glorified TiVo, what am I using to replace last week’s wine-besotted and (therefore) prematurely defunct system (whose HDD, thankfully, is still fully functional)? In an example of the ‘when life hands you lemons, make lemonade’ strategy, I’ve further explored the VMware Fusion-on-MacBook experiment that I began back in late May when my Boot Camp-created partition went AWOL when I was overseas. As such, I’ll be able to supplement the conceptual discussion of software virtualization planned for my October 2 cover story with some real-life benchmarking results and day-to-day operation observations. And, through my to-date testing that began last Friday evening when I started installing and running software on the 20 Gbyte virtual NTFS partition, you can call me quite impressed so far.

Nothing that I’ve yet tossed on this virtualized Window XP SP2 (originally, upgraded to SP3 plus patches) O/S-plus-applications image, including the entire Office 2000 suite, has given me even a halfway-strong suggestion of a grumble. The only notable glitch that I’ve uncovered so far, in fact, is that I was originally unable to ’see’ other LAN resources on my workgroup, until I switched the virtualized networking connection from NAT to Bridged mode (NAT mode is nice, for example, when you’re in a hotel that bills Wi-Fi access on a per-MAC-address basis, because the virtualized O/S doesn’t then appear to the hotel network system as a distinct ‘computer’). USB-tethered HDDs and flash drives work fine.

By extracting requisite drivers from the Boot Camp 2.0 image on the OS 10.5 CD, I’ve even been able to get the MacBook’s built-in webcam working in virtualized Windows…and although I haven’t yet tried, I suspect that getting the computer’s IR and Bluetooth facilities operational in the VMware-fueled O/S would be equally easy. ActiveSync connections to my Windows Mobile Smartphone ’should’ be similarly straightforward. And, by relying on equivalent applications included with Mac OS 10.4 or installed under it on my comparatively voluminous HFS+ partition, I’ve been able to keep a surprisingly high percentage of my virtualized 20 GByte NTFS partition unoccupied…especially if I remember to regularly compress it!

I’ll save the virtualized-vs-Boot Camp native benchmarking results for my 10/2 print writeup, but for now I’ll just say that virtualized Windows XP on this particular piece of hardware feels at least as speedy as when run native on its Dell Inspiron 700m predecessor, maybe more so…and it’s rock-solid to boot (certainly more so than with my past Boot Camp-on-MacBook experiments). Frankly, though, rarely do I kick off an operation that would previously stress the CPU in native-Windows mode and therefore would completely max it out in its currently OS 10.4-plus-VMware Fusion-plus virtualized Windows XP configuration. With that said, the system fan is spinning more frequently and more raucously than when I’m running OS 10.4 alone, or for that matter in this system’s Boot Camp past history, thereby suggesting that the CPU is on average substantially more heavily loaded in its virtualization-augmented present state. And battery life predictably…consequently…notably suffers as well.

Note, too, that I’m not doing anything 3D-graphics-intensive, as in-production VMware Fusion v1 offers only ‘experimental’ support for Direct3D v9 (an option which I haven’t even bothered enabling). Just-released Fusion v2 Beta 2 promises to improve graphics subsystem virtualization speed and feature robustness, as well as to bolster other processing-intensive functions such as CPU-centric video decoding. Hopefully, by the time my article appears in print, Fusion v2 will exit beta and I will therefore be able to pass along my updated benchmarking and functional observations via addendums on the Brian’s Brain blog.

Final comments, for now:

• Right now I’m running VMware Fusion in a mode in which it (if necessary) can harness both of my CPU’s cores. This, in fact, is the crux of why I migrated to VMware Fusion from Parallels Desktop, which currently can only tap into one core’s resources. With that said, on average I noticed little to no performance difference in Windows XP between Fusion’s single- and dual-core operating modes.
• I also bumped up the amount of system memory available to the virtualization scheme to 1 GByte. I daresay this optimization probably has a great deal to do with the resultant impressive perceived Fusion performance, by reducing the probability that Windows will need to employ the virtualized HDD partition as swap space…but were I to be in parallel doing anything substantial memory- and/or performance-wise under OS 10.4, I might have left the virtualized memory setting at its 512 MByte default.
• Power management, by leveraging OS X’s foundations, is such a joy as compared to times past. When I want to put the system to sleep, I close the lid or punch the power button. Everything goes into doze, OS 10.4-native and Windows XP-virtualized alike. I lift the lid, and everything comes back to life without a hitch. Earlier today, for example, I was ‘virtually’ streaming an album off the Zune Marketplace within Windows and, as a test, I put the system to sleep mid-song. When I woke the MacBook back up later, the tune immediately continued without missing a single beat. Impressive.