This seems to come up every few years, so I figured I'd revive it and see
if anyone has better answers now. :)
I've got several 6509s with WS-X6408A-GBIC and WS-X6416-GBIC blades where
I'm seeing output drops on GigE interfaces that (according to 30s avg) are
typically doing just a few hundred mbit/s...i.e. anywhere from 20-50% line
rate utilization and I'm seeing thousands of output drops per day.
There are no interfaces larger than 1gbit, so it's not like there are very
short lived microbursts coming from 10g ports swamping the 1gb port
buffers.
No amount of fiddling with output hold-queue size, wrr-queue bandwidth, or
wrr-queue queue-limit has eliminated the drops.
I realize these blades are bottom of the barrel as far as gigabit ethernet
goes on the 6500. But even the newer cards don't have a whole lot more
buffer space per port. Our cards have 512k per port. WS-X6516A has 1MB.
The WS-X67xx cards have 1.17MB TX/166KB RX buffers.
Would swapping out the WS-X64xx cards for WS-X6516A's (with or without
DFC3BXL) likely make much difference?
----------------------------------------------------------------------
Jon Lewis, MCP :) | I route
Senior Network Engineer | therefore you are
Atlantic Net |
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
_______________________________________________
cisco-nsp mailing list [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
