[c-nsp] 3560 buffering

Peter Rathlev peter at rathlev.dk
Wed Oct 14 06:05:37 EDT 2009


On Wed, 2009-10-14 at 10:45 +0100, Mateusz Blaszczyk wrote:
> On Tue, Oct 13, 2009 at 05:59:55PM +0200, Peter Rathlev wrote:
> [...]
> > ! Only output queue-set 1, queue 2 is used. Adjust all thresholds to
> > ! 400% of default. (This is AFAIK the maximum even though the parser
> > ! accepts up to 3200%.)
> > mls qos queue-set output 1 threshold 2 400 400 100 400
> 
> So main differnce to "no mls qos" is that thresholds are lifted from
> default 100% to maximum 400% - is that correct - meaning the packets
> are dropped later then earlier?

Default thresholds can be seen with "show mls qos queue-set (1|2)";
AFAICT the defaults are 100%. So yes, the command has the effect of
allocating more buffer space, meaning less drops on bursts. I still
don't quite understand why the default ("no mls qos") won't use all
buffer space, but our lab tests points towards that they simply don't.

> > ! Assign all buffers to queue 2 (also used by the CPU)
> > mls qos queue-set output 1 buffers 0 100 0 0
> > !
> 
> What is the default here? All queues are used by default?

Hm... looking at a switch that has never had "mls qos" configured:

someswitch#sh mls qos
QoS is disabled
QoS ip packet dscp rewrite is enabled

someswitch#sh platform port-asic stats enqueue gi0/1

  Interface Gi0/1 TxQueue Enqueue Statistics
    Queue 0
      Weight 0 Frames 2
      Weight 1 Frames 0
      Weight 2 Frames 0
    Queue 1
      Weight 0 Frames 0
      Weight 1 Frames 34736
      Weight 2 Frames 318358119
    Queue 2
      Weight 0 Frames 0
      Weight 1 Frames 0
      Weight 2 Frames 0
    Queue 3
      Weight 0 Frames 0
      Weight 1 Frames 0
      Weight 2 Frames 425983701
someswitch#

It seems that all queues are actually used according to the default CoS
map. I think I'm getting confused here. Can anybody shed light on this?

The reason I used queue 2 in the example is that the CPU apparantly uses
it so you can never allocate 0% for this queue. The "no QoS" scenario is
best achieved with one simple queue using all buffer space.

-- 
Peter




More information about the cisco-nsp mailing list