[c-nsp] WS-X6348-RJ-45 and Output drops
Peter Rathlev
peter at rathlev.dk
Thu Feb 7 10:35:30 EST 2008
On Thu, 2008-02-07 at 09:41 -0500, Jon Lewis wrote:
> On Thu, 7 Feb 2008, Robert Hass wrote:
>
> > Can I configure something to prevent this drops ? Or maybe I should
> > replace this card with WS-X6148 will resolve my problem ?
>
> According to cisco docs, the per port buffers on the 6148 are no
> better. WS-X6548-RJ-45 has 1MB/port buffers. We had a 6500 with a
> busy FE port, and swapping from a WS-X6348-RJ-45 to a WS-X6548-RJ-45
> made a big difference in the output drops (but didn't totally stop
> them).
Right about the WS-X6148-GE-TX, but if you rather want a bus card the
WS-X6148A-GE-TX (note the "A") has 5.5 MB buffers per port and 1p3q8t
queuing with WRED, DWRR and even SRR if your supervisor/IOS supports
it.
The WS-X6148A-GE-TX is oversubscribed though, having one Gigabit ASIC
per 8 ports. Here the WS-X6500 cards are better.
On Thu, 2008-02-07 at 13:54 +0000, Robert Hass wrote:
> Can I configure something to prevent this drops ? Or maybe I should
> replace this card with WS-X6148 will resolve my problem ?
<snip>
> queue thresh cos-map
> ---------------------------------------
> 1 1 0 1
> 1 2 2 3
> 2 1 4 5
> 2 2 6 7
<snip>
> Packets dropped on Transmit:
> BPDU packets: 0
>
> queue thresh dropped [cos-map]
> ---------------------------------------------------
> 1 1 9839111 [0 1 ]
> 1 2 0 [2 3 ]
> 2 1 0 [4 5 ]
> 2 2 0 [6 7 ]
<snip>
You can select what the switch should rather drop. Right now you
probably have everything classified as CoS 0 and equal chances of
dropping. With more diverse classification and scheduling you can
decide types of traffic that shouldn't be dropped.
For general info on configuring QoS on Cat6500, look at:
http://www.cisco.com/en/US/docs/switches/lan/catalyst6500/ios/12.2SXF
/native/configuration/guide/qos.html
(http://tinyurl.com/2y725o)
Take a look at this for information about buffer sizes etc.:
http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps708/prod
_white_paper09186a0080131086.html
(http://tinyurl.com/36c57a)
Regards,
Peter
More information about the cisco-nsp
mailing list