Hi everyone,
This one really has got me baffled. Our Lite Touch image is ~200MB, and takes about 1 minute and 15 seconds to download from WDS if one client is booting the WIM.
The moment I do two clients, it goes up to about 2 minutes (and so on and so forth).
Regardless of this, it seems our WDS server has set a cap of 0.1% of the available bandwidth for TFTP.
The server itself is an HP DL180G6 with 4 gigabit NICs all teamed in 802.3ad Link Aggregation using the HP teaming software. The clients are mostly HP dx2450 machines with NVIDIA gigabit bootable NICs.
I can't seem to find a way to persuade WDS that it can use more than 0.1% of the bandwidth.
We have adjusted the TFTP block size, and any more than 1422 causes problems (we lack jumbo frame support in places!), and we've tried TFTP window sizes from 1 through 16 - no performance increases (many decreases at higher numbers!).
We've even been through http://blog.danovich.com.au/2011/07/28/slow-sccm-osd-tftp-pxe/ as well, and upped the buffers on each card to 2048 (although we can't seem to adjust them on the team itself in the HP software!)
Has anyone got any ideas? This is driving us up the wall!
Cheers,
Chris