Letters

OpenOffice and Terminal Services
I was just reading Jeff James’ review of OpenOffice 3.0 ("OpenOffice 3.0 Challenges Microsoft's Office Dominance," December 2008, InstantDoc ID 100545). I've been considering OpenOffice as a replacement for Microsoft Office 2003. However, I've been unable to find any information about how the product performs in both Terminal Services and VMware ESX environments. I'd love any information you can provide.
—Dave Warnes

You shouldn't have any trouble virtualizing OpenOffice 3.0, but I've heard mixed reports/results about using OpenOffice with Terminal Services. For example, check out the comments that follow the article "Terminal Services Plus OpenOffice Equals" (lesmurphy.com/2009/01/25/terminal-services-plus-open-office-equals). An article from our site that you might find handy is "OpenOffice Registry Fix" (windowsitpro.com/article/articleid/93970/openoffice-registry-fix.html). Finally, the OpenOffice.org support forum (support.openoffice.org) might also be helpful.
—Jeff James

Windows 7 Test Drive
After reading Michael Otey's recent article “Windows 7 in the Enterprise” (June 2009, InstantDoc ID 101885), I took the Windows 7 Release Candidate (RC) for a test drive on my Dell Latitude D600 laptop, currently running Windows XP SP3. The system has a 1.7GHz Intel processor, 1GB RAM, an ATI Mobility Radeon 9000 video card, a Broadcom 570x NIC, and an Intel Pro Wireless LAN 2100 3a Mini PCI adapter. (A few years ago, I tried upgrading to Windows Vista 32-bit but couldn’t find support for the hardware components.) I use this relatively old computer for Microsoft Office, remote controlling other computers with the RDP client, and running guest machines under Virtual PC 2007. My only concern is an increasing slowness as more and more Microsoft patches are applied.

The promise of a more nimble, lighter-footprint OS from Microsoft intrigued me, so I downloaded the Windows 7 RC. I burned the download to a DVD and installed a spare disk drive on my laptop just to be sure I wouldn’t have to go back to square one if the test didn’t work.

The installation went relatively fast, and the laptop booted quickly. The Microsoft standard VGA drivers worked well, and I had a functioning wired Ethernet connection. I needed sound, so I installed Sigmatel audio drivers from Dell. Before I work on a new OS installation on a laptop or workstation, I get the latest drivers from the manufacturer’s website and compile them on a CD. If the NIC doesn’t “light up,” I can install any drivers on the spot. I was unable to find any drivers for the ATI video; however, the Microsoft-supplied drivers were good enough.

My real problem was the Intel Pro Wireless 2100 3A adapter. A notebook without wireless capability just wouldn’t do. I found drivers for Vista on the Intel download site installed them and connected to my home network. At last, I had a fully functioning computer! Now for the real test: I needed antivirus, so I downloaded Avast! Home edition and then installed Microsoft Office 2007 and Virtual PC 2007.

While I’m not running in an enterprise environment, I find the Windows 7 startup to be surprisingly quick and application performance to be adequate to my needs. Also, if you’re familiar with Vista, the UI isn’t much different, and the UAC seems to be reasonable with its default setting. All things considered, Microsoft may be on to something.
—John Swanson

VMware ESX 3. 5 and Processor-Virtualization Changes
I read John Savill’s FAQ, “Does VMware ESX 3.5 require a 64-bit processor with hardware virtualization features?” (June 15, 2009, InstantDoc ID 102301). John says “ESX 3.5 doesn't currently take advantage of the hardware assist technologies (Ring -1) in the Intel and AMD processors. VMware uses binary translation, which they have found gets better performance than the native hardware virtualization in processors.” The understanding is that future versions of VMware will utilize some of the hardware-virtualization assistance.

Actually, ESX 3.5 does. In fact, VMware was one of the first virtualization vendors to recognize hardware I/O virtualization, with support for Intel-VT, AMD-V, N Port ID Virtualization (NPIV) on Fibre Channel cards, and switches and nest-page-tables for memory. Furthermore, enabled Intel-VT—which supports long-mode processors—is actually a requirement to get a 64-bit OS to work on ESX 3.5. In reality, VMware uses a combination of Direct Execute (the cause of CPU-compatible requirements for Vmotion, aka live migration) and binary translation for virtual interrupts (in the main networking and disk access). VMware also supports paravirtualization if the guest OS is compiled appropriately, as with Fedora Linux.

Finally, for some time, it’s been possible to virtualize ESX on ESX and VMware Workstation. Doing so requires editing a configuration file (.vmx) to open a backdoor, but afterward you can run ESX on ESX, and even run a virtual machine (VM) on the “virtualized ESX” VM—pretty crazy stuff, but handy for people who want to test the VMware products with limited hardware resources.
—Mike Laverick

I checked with the VMware engineers, who responded, “Both answers are somewhat correct. VMware doesn't use Intel-VT and AMD-V in ESX 3.5, except for 64-bit VMs. VMware’s binary translation was found to be faster for most workloads than Intel-VT or AMD-V. Hyper V and Xen require Intel-VT/AMD-V for all VMs as they leverage the VM monitor (VMM) that Intel-VT and AMD-V provide.” I've updated the FAQ to point out that the processor assist is used for 64-bit VMs. Thanks for bringing this to my attention.
—John Savill