Reduce OSD Time

, , ,

Reduce OSD Time

1. Set High Performance Power Plan

  • This has proven in multiple environments that I have been to save 20% – 50% on certain steps, such as download/apply the OS. Also note that by default configmgr while in WinPE will run power setting set to balanced, so making sure we use all resources we can see a major win from this piece. I also set this to prvent systems from going to sleep during the TS.

2. Set SMS Host Agent start-up options

  • By default the SMS Host Agent (CcmExec) service is set to a delayed start. So every time you restart the system during the OSD Process and you see the “initializing System Center Configuration client” you are just costing time. In my environment this message would usually last 2 – 5 minutes each restart. Once I implemented this we noticed the configmgr client would only take < 15 seconds to initialize, that is a savings of 800% – 2000% for this specific task sequence action.
  • Frank Rojas whom is one Microsofts CSS Lead OSD Engineer has let me know that this has POTENTIAL for failures in a TS. So please know that if you use this step please understand it is not supported by Microsoft. I have not personally seen any problems with this, but I do understand there is possible for OSD failure as a result.

3. Set SMSTSRebootDelay

  • This will change the default behavior from 30 seconds to 0 seconds and occur immediately
4. Set manual Restart times (for displayed messages)
  • By default when you add a restart computer step into the TS the time is set to 60 seconds. Simply by un-checking the “Notify the User before restarting” box you now force the restart immediately. Should you need to actually display a message you can just reduce the amount of time this box is displayed.
5. Eliminate unnecessary restarts / steps
  • I have seen several customer environments that had 8 or more restart steps in their TS. Whenever you have a restart in a TS you can assume usually 2 – 5 minutes are added to the total TS time. I would recommend investigating the status messages for exit codes indicating a restart is necessary before you start modifying a customers task sequence. Many of my peers usually have only 2 or 3 restarts in their TS as they have streamlined their deployments.
  • MVP Mick Pletcher has a great post on how to handle restarts in your TS. Basically add a condition statement to check if the system needs a restart you can read about it and implement it here
6. Patch the WIM
  • This will of course help you out on the install software update steps in your TS. The more patches that are in your WIM, the less to process later. I tend to bake office and patches into the WIM. I’m not a fan of having really thick images loaded with several 3rd party apps that can be installed during ts.
7. Verify boundaries are efficient
  • This should be another no-brainier idea, but you will be surprised. I’ve seen customer environments where a machine from the east cost USA would pull down images from the middle east….like there is literally an entire ocean btwn the DP and the workstation..
8. Prevent OOBE From connecting to Windows Update
  • I’ve seen in the SMSTS.LOG before that a system I was imaging was trying to reach out to windows update. Some places you are confined to image from certain vlans only and you can block this communication at the firewall. In my TS I modify the Registry to prevent this action from occurring. Unfortunately this was something changed in my TS a long time ago so I did not grab a screenshot. I can edit this post when I get around to making the next revision of changes in the updated WIM. (Thanks Todd)

On my test experiment the OSD process would take around 100 – 110 minutes…that’s nearly two hours, and wayyyy too long.

Reduce OSD time


I like to use Thomas Larsens OSD Dashboard when I keep track of some Task sequence information like this. Please check out his blog here. I believe this dashboard is a must have for every SCCM admin.

Screenshot after changes OSD Time. We see below that this has been reduced to 36 minutes. I am positive I can actually get this to 32 minutes or less, but for now I will leave good enough alone. This test shows that we have cut the OSD time to less than 1/3 of the time. I’ve been in other environments where they had the image process take 24+ hours, but that is mainly boundary related issues.




Power settings: I kind of have this littered everywhere in my TS before major time consuming steps like apply OS, install patches, install apps, etc.

Name: Set High Performance Power Scheme
Command Line: PowerCfg.exe /s 8c5e7fda-e8bf-4a96-9a85-a6e23a8c635c
SMS Host Agent Start-Up Properties: This only needs to go in 1 place right after you install the SCCM Client. 

NameSet SMS Host agent to start immediately
Command Linecmd /c sc config “CcmExec” start= auto

Dynamic Variables: I have all my variables set here. These are for resiliency, and speed.


Prevent OOBE from Downloading Drivers:

NameMount the Offline HKLM File
Command Linereg load HKLMOffline C:WindowsSystem32ConfigSoftware

NameSet DODownloadMode
Command LineREG ADD HKLMOfflineSOFTWAREPoliciesMicrosoftWindowsDeliveryOptimization /v DODownloadMode /t REG_DWORD /d 100

NameSet DoNotConnectToWindowsUpdate
Command LineREG ADD HKLMOfflineSOFTWAREPoliciesMicrosoftWindowsWindowsUpdate /v DoNotConnectToWindowsUpdateInternetLocations /t REG_DWORD /d 1

NameUnmount the Offline HKLM File
Command Linereg unload HKLMOffline

End of TS to change back to default (if wanted)


NameSet SMS Host agent to start delayed
Command Linecmd /c sc config “CcmExec” start= delayed-auto

Name: Set Balanced Power Scheme
Command LinePowerCfg.exe /s 381b4222-f694-41f0-9685-ff5bb260df2e


This customer has unnecessary restarts in their OSD design, and were not very streamlined. Things in this version of the TS can be easily baked into the WIM on the next image refresh with the build/capture TS i provided for the customer. There are things that are slow during this TS like install Visual J+, or .Net Framework 3.5 that would reduce the OSD time if baked into image. 

Other things to reduce OSD Time: 
I still have unnecessary install software update/restart steps in the TS as a sort of catch all. These systems come out fully patched, but are left enable in case I do not get time to capture and test a new wim one month.

I have seen in the SMSTS.LOG where systems would try to reach out to windows updates for patches instead of staying within my environment. This can be changed via registry modification…I have this in place in my environment but did not capture screenshots when I saw it in the log files. 

Successfully added DaRT to boot image….or did it?

, , , , , , , , ,
Successfully added DaRT to boot image….or did it? Here is how to identify the problem and  a link to fix it!
I was recently onsite with a customer where the proposed design document included MDOP DaRT integration into the boot images. DaRT is a great tool to have because it gives the engineer the ability to remotely connect to the machine while within the WinPe environment. This particular customer is undergoing a massive and understaffed windows 10 migration where every bit of efficiency really makes a difference on deployment nights.
First a quick review on installing MDOP DaRT, Enabling Monitoring, and creating the boot image.
  1.  Install MDOP DaRT on primary site server
  2.  Copy the Toolsx86/64 cab files into proper directories into the MDT deployment share
  3.  Enable Monitoring on deployment share
Deployment share \SERVERD$DeploymentShare
Ports: 9800 (Event port) 9801 (Data port)

Connect to deployment share > Right click on “Monitoring” > Navigate to Monitoring Tab and fill the check box

Once this is filled you will start to see systems as they image from this view. 
If you are in an environment that is not really using the MDT deployment share you would still open up the MDT toolkit and modify the CustomSettings.INI. This customer is heavily utilizing the MDT Deployment Share with all the settings applied we can access the “Rules” tab and see the setting is automatically applied after we enabled monitoring. The great part about using the deployment share in this scenario is that we can make constant on demand changes and not have to worry about hash mismatch errors like if were working within the MDT toolkit package.
 We are now able to make our DaRT integrated boot image from the console on our primary site server. Begin by selecting “Create Boot Image using MDT” Make sure to select the following optional components “MDAC/ADO Support, and DaRTT”
From this point we distributed the enabled the boot image for PXE deployment, added drivers, and attach it to a task sequence. In the screenshot below you will notice we are missing something? We do not have the “DaRT Remote Control” option that we should have.


NOTE: Sometimes when the boot image is “Successfully” created it does not add the “DaRT” tool. I am able to verify this to be a LIE by looking into the PEMananger.LOG located in my temp folder.


When we look at the command that was ran by accessing the “RunCMD.CMD” we see that only the WinPE-MDAD_EN-US.CAB is the only package even attempted to be added.
You can investigate further by opening up DISM GUI and searching for any trace of DaRT on the boot image. As you can see DaRT did not even attempt to be installed into the wim.
Manually modify boot image to include Dart functionality by using the script below.
HOW TO FIX IT: Johan Arwidmark has a script available online that I have used to inject the Dart into a newly created WIM.


Once we ran the script created by Johan and injected the drivers I was able to start using DaRT tools.
After the USMT toolkit is called and the Gather step starts to run a box on the bottom left will appear  on the system being imaged but minimized. This is your indicator to let you know that you can now use DaRT functionality.


From the Monitoring Node in the deployment workbench right click the computer we are trying to troubleshoot > Select Properties > Select DaRT Remote Control
Do not always take the console UI at face value and always verify with log files. Some occasions the console indicates something was done correctly but you need to check the logs. If this happens then you need to go old school and use the tried/true methods. If you run into a problem always do a quick search b/c the Deployment Research guys might already have a work-around.
To vote for this to be fixed from SCCM team please visit the link below.