[c-nsp] Excessive output-drops on WS-X6224 module on Cat6500

[Steve Fischer]

We have had a performance issue that has been plaguing us for some time.  Our access switches are 2924M's, which have given us difficulties in a variety of fashions.  These converge at a Catalyst 6500 w/redundant Sup2's+MSFC2's.  As these 2924's aren't even pushing 1mb sustained, I'd have not much reason to suspect the 6500.  We, due to their lesser nature, and the various other issues we'd had with them, always assumed the performance issues were related to the 2924's, but, upon detail inspection of the ports on the 6500, we are seeing a boatload of Output drops on the (FastE MTRJ) interfaces that are feeding the 2924's.  These FastE modules in the 6500 are WS-X6224 - we have other 6500's with the WS-X6324's that don't seem to be exhibiting the same symptoms of excessive output drops.  

1. Has anyone else experienced this issue?

2. Cisco Engineer has told us (ok...I inferred the oversubscription part of this) that there but 64K out buffer per port to be divided between input and output - this could be related to the buffers getting 
overrun.  He also indicated that there are (2) 512K buffers per WS-X6224 module that are inactive until they are turned on with an undocumented command : wrr-queue group-buffers  on the first port in the group of 12.  Has anyone ever used this command/auxiliary buffer space before?  What were the results?  Is it reasonable to expect this to resolve my issue?
Steve Fischer
sfischer at myrealbox.com