I’d like to start this post with a statement:

Windows Deployment Services (WDS) is a useful role in Windows Server 2008 R2.  It’s free (to licensed Windows users), supports multitasking, and is a perfectly good method of pushing Windows images to clients…

Unfortunately that statement has a caveat:

… but it needs to be installed on an Active Directory-member computer.

For some, that’s a non-starter.  And sometimes, you just want a quick and dirty solution.

I have a small dedicated server at home to run Active Directory along with basic network services (DNS, DHCP, etc.) for my home IT.  I also run Philippe Jounin’s excellent TFTP Daemon (service edition) on it in order to support image loads on my Cisco 7940 IP Phone.

In order to rebuild the Hyper-V server that I use for infrastructure test and development, I wanted to boot across the network and install Windows Server 2008 R2 – and a few days ago I found Mark Kubacki’s post about TFTPd32 and DHCP Server – Windows Deployment Services without WDS. Perfect!  No need to install another role on my little Atom-powered server – particularly as, once this server is built, I’ll probably install WDS on it  to deploy images to my various test virtual machines!

So, this is the process – with thanks to Mark Kubacki, and to Ryan T Adams (who wrote about installing Vista without a CD Drive using TFTP – for instance, installing Windows on a netbook) who were gracious enough to blog about their experiences and give me something to build upon:

1. Download tftpboot.exe from Ryan T Adams’ site and run it to extract the contents to a suitable hard drive location (i.e. the TFTP root folder).  Unfortunately, you probably won’t need most of this 154MB download (more on that in a moment) but it will get you started.
2. Start tftpd32.exe (or copy the files to your TFTP root, if you are already running a TFTP service, as I was) and add tftpd32.exe (or tftpd32_svc.exe) as a Windows Firewall exception (you could just disable the firewall but I don’t recommend that approach).
3. Either set TFTPD32 to act as a DHCP server and specify the boot file options (as Ryan describes), or configure DHCP options 066 and 067 (boot server host name and boot file name) on another DHCP server (Mark shows how to do this for the Windows DHCP Server role) using the IP address of the TFTP server and the boot file name of boot\pxeboot.com.
4. Make sure that the TFTP Server is set to include PXE capability in the advanced TFTP options and that it’s DHCP Server capability is turned off if you are using another DHCP server.
5. Restart the TFTP Server (or service) to pick up the configuration changes.
6. Boot a computer (or virtual machine) from its network card, press F12 when prompted and wait for Windows PE to load, then map a drive to another machine on the network which is sharing the Windows media (I use Slysoft Virtual Clone Drive to mount an operating system ISO file and I’ve shared the virtual drive).
7. Switch to the newly mapped drive and type setup.exe to run Windows Setup.

Unfortunately, the version of the Windows Preinstallation Environment (Windows PE) that Ryan has supplied in tftpboot.exe is a 32-bit version (of Windows PE 2.0, I think).  When I tried to use this to install Windows Server 2008 R2 (which is 64-bit only), I was greeted with the following message:

This version of Z:\setup.exe is not compatible with the version of Windows you’re running.  Check your computer’s system information to see whether you need a x86 (32-bit) or x64 (64-bit) version of the program, and then contact the software publisher.

I needed a 64-bit version of Windows PE.  No problem.  That’s included in the Windows Automated Installation Kit (WAIK), so I overwrote Ryan’s winpe.wim with the one from %programfiles%\Windows AIK\Tools\PETools\amd64, and restarted the computer I wanted to build.  This time Windows Setup ran with no issues and Windows Server was installed successfully.

Even though I used the TFTPD32, this method could be used to install Windows from just about any TFTP server (it could even be running on totally different operating system, I guess), or even to load another WIM file (i.e. not Windows PE) from a network boot. I’m sure if I had more time I could come up with all sorts of scenarios (boot Windows directly from the network?) but, for now, I’ll stick to using this method as a WDS replacement.