Generic QoS question

Unanswered Question
Mar 7th, 2007

Could someone tell me if a QoS policy map affects traffic only in a congestion situation, or is QoS constantly being applied to traffic regardless of utilization? Thanks.

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 4.8 (5 ratings)
Loading.
n-shah Wed, 03/07/2007 - 19:44

Hi There

QoS policy (shaping, policing) kicks in only when there's congestion.

However, classification & marking happen regardless of congestion.

rgds

nick

jarvar832004 Wed, 03/07/2007 - 23:19

QOS is basically congestion management tool, so there is really no use of it when there is enough bandwidth for all traffic. The difference can be seen only when the bw is occupied with different kinds of traffic when the highly marked traffic gets prioritised.

mheusinger Thu, 03/08/2007 - 00:00

Hi,

just some clarifications.

QoS tools available:

1) Classification

2) Marking

3) Policing

4) Shaping

5) WRED

6) Queueing

7) Compression (L2 header or payload)

8) Link Fragmentation and Interleaving (LFI)

WRED and Queueing are the only two, which depend on congestion. The rest (especially policing and shaping) are applied to all matching traffic, regardless of congestion.

Policing and shaping are typically used in a service provider environment to enforce subrate access. Shaping f.e. is also helpful with EoSDH, f.e. provider installs a GE interface to the customer, but in reality a STM4 (622 mbps) is used to transport data. You better shape the GE down to 600 mbps then on the customer equipment.

Congestion management is in many cases the dominant feature, but not everywhere. As already stated in the previous posts, this feature kicks in, when congestion happens. Queueing configuration lets us sort out, which traffic should be forwarded first then and which traffic could wait a little.

Compression and LFI are used on small bandwidth links to get more throughput or address jitter issues with voip traffic.

If you need detailed technical informations, I would recommend to read through the Enterprise QoS Design Guide found at www.cisco.com/go/srnd

Hope this helps! Please use the rating system.

Regards, Martin

wmelickbts Thu, 03/08/2007 - 14:27

Thanks. I think this clears it up for me. I have Cacti set up to show drops of some of my output policy-maps (I am marking packets with DSCP for MPLS). Occasionally I see drops even though the line itself seems to be only 30% used. However I think that my 5 minute sampling interval may be hiding some short-lived traffic spikes where drops may occur.

sundar.palaniappan Thu, 03/08/2007 - 14:36

On the router, you can configure 'load-interval 30' under the interface to get a better picture of the interface utilization. This would show the 30 second input/output rate of the interface and that's the closest you can get to real time utilization.

HTH

Sundar

Actions

This Discussion