Time to deploy Windows 7. You take a look at your desktop and laptop inventory. Is it going to be easier to create an image for each model, or to create one image and to add the different drivers and components required for each model? Adding drivers sounds more efficient, but creating different images is more straightforward. What are the trade-offs? And what tools do you need? The more you think about it the less clear it can seem.
If you have a limited range of models, and you decide to create one image for each model, you can use Windows Deployment Services (WDS) alone to create and deploy the image. You will:
- Build your reference image
- Run Sysprep to generalize it
- Capture it
- Add an unattend.xml to automate the deployment
- Deploy it to the same make and model of computer.
However the default sysprep /generalize command during image capture strips out computer-specific information. This includes the device detection and driver selection information. In the sysprep /specialize pass during image deployment the computer re-runs the device detection and driver selection, and re-installs the drivers.
This adds minutes to each deployment. It can be a few minutes for a simple VM image, up to twenty minutes or more for a laptop. You can save a lot of time by using the sysprep option PersistAllDeviceInstalls in your Unattend to avoid stripping out and reloading the same drivers.
Here is the Unattend:
<unattend xmlns="urn:schemas-microsoft-com:unattend">
<settings pass="generalize">
<component name="Microsoft-Windows-PnpSysprep" publicKeyToken="31bf3856ad364e35" language="neutral" versionScope="nonSxS" processorArchitecture="x86">
<PersistAllDeviceInstalls>true</PersistAllDeviceInstalls>
</component>
</settings>
</unattend>
This has to go into an Unattend file used during image capture, not image deployment. Using WDS you will need to add the custom unattend.xml file to the reference image and run sysprep at the command line instead of the GUI before shutting down and capturing the image.
Customizing the Generalize pass in WDS
This command line translates into the following:
- Run the Generalize pass now
- Run the Out of Box Experience when the computer next starts
- Shutdown instead of restart, so the image can be captured with WinPE
- Use the provided xml
This is just a small tweak to your workflow when using WDS to deploy images to the same make and model of computer. Don’t forget, you need to sysprep and capture an image that has not been joined to the domain.
In Microsoft Deployment Toolkit (MDT) 2010 the process of creating the reference image and capturing it is automated in a Task Sequence. However there is no opportunity to interrupt the task sequence to change the unattend.xml. Instead you modify the unattend file for the task sequence beforehand in the Workbench.
Select the OS Info in the Task Sequence
Edit the unattend.xml and save it
There is an option in the WDS Client (aka Windows Deployment Wizard) to Prepare to capture the machine, but stop before running sysprep. This would give you an opportunity to customise the image, including the unattend file.
But there is no option to resume the capture, so you would choose this option if you want to hand the image over to a different capture process.