MDT Server demands excessive memory when deploying

[Philip60]

Hi All,

I've set up a Hyper-V W2012 server to deploy images, configuring it to use dynamic memory. When I deploy images, (using MDT2012) they start off ok, but slow down to a snail's pace (eg. the whole process takes 4 hrs+), and the server's memory requirement goes up and up, currently demanding 13,757MB to deploy 5 images! OK, they are quite 'fat' images (the wim is 6GB) but even so... Any ideas?

Thanks.

[MordyT]

And we deploy 10gb wims... 25 concurrent in unicast.... With 4gb of ram....

 

Mdt2010 here.

[free780]

What is the server os for mdt? Which process uses the memory when it peaks? Are you using multicast?

[Philip60]

1. 2012 2. Erm, not sure - will have to do another deployment to find out! 3. Yes Thanks.

[morganw]

MDT deployment is client driven, with the deployment share just being a regular share (you can run it from a Samba server if you wish). If memory usage is that bad I would imagine that this is WDS running the multicast and encountering problems. The process name (question 2 above) should confirm this and if you can deploy 5 machines with unicast without issue, it's probably an IGMP issue on a switch.

[Philip60]

OK, many thanks, I'll check it out.

[computer_expert]

If you are using the database portion of MDT, it may be that SQL server is using the majority of the ram.

[morganw]

The database will be small, and the only connections are probably in the Windows PE phase and a one-time read. If unicast deployment is slow as well, then I would look for performance problems with the VM, e.g. TCP Chimney Offload.