I was recently involved in a project were it was decided that a “thick” image should be used, so the image ended up to be 94gb (!) It was also decided to use MDT 2012 to do the deployment. When I did the test runs at home in my lab all went well and no problems occurred. So the next step was to implement the deployment share at the customer, everything went smooth and we decided to use Multicast so we didn’t kill the network. We had two test computers, one with a 256 gb spin disk and one with a 160 gb SSD.
First we tested to deploy the one with the spin disk and we got a successful deployment. Customer was happy an I was happy. Then just to be shure and for benchmarking we started the deployment to the SSD machine, BANG…
My first thought was that the User State had kicked in and stored the data locally, but it had not. So the next suspected was the Multicast transfer. I suspected that when you use Multicast, MDT downloads the whole image locally and then starts the installation. I contacted my colleague Olav Tvedt and asked him about how Multicast handles the .wim file and he confirmed my suspicion. The image is compressed and transferred to the disk with a 10% buffer, then starts the installation. So with some basic math you can see that with multicast you will run out of disk space.
Thank you Olav for a quick answer, as usual 🙂