Running SuSE SLES10/SP2 and an image of about 17GB. When multicasting clients, performance appears to suffer from a excessive number of timeouts and/or retransmits. If the images are smaller this is not a problem, but we currently need them this size.
The errors appear frequently:
Timeout notSnasweed=[1] notReady=[1] nrAns=2 nrPart=3 avg 10193
We added the configs of rmem_default and rmem_max in the /tftpboot directory substructure in init.d for the client. It does seem to affect the speed, but still winds up bombing out on the big partition (~17GB).
It's an IBM blade center w/SAS drives which do not have cache.
TIA,
Brent