WAN multi-site traffic shaping/QoS setup

[Scott Bailey]

We have multiple sites connected to a private WAN cloud (Layer 2) with varying CIR's.  Our QoS setup from our main site is 3 tiered in order to shape traffic to our overall CIR at the main site and shape traffic on different child classes based on the remote CIR.  Parent policy is applied to interface connected to the service provider.  We continue to have drops even though we don't seem to be reaching the CIRs for the remote sites or the CIR limit on the ciruit.  I've worked with TAC and we have adjusted queue limits based on drops but this is an ongoing issue.  Should there be a relationship between the queue limits and shape averages?

Sample config
NOTE: Shape averages are 95% of CIR
Policy-map Parent
class class-default
  shape average 142500000
  queue-limit 1024 packets
   service-policy ASE-Remotes-Policy
policy-map ASE-Remotes-Policy
Class site1
shape average 4750000
  queue-limit 400 packets
   service-policy ASE-QoS-Policy
Class site2
shape average 19000000
  queue-limit 1024 packets
   service-policy ASE-QoS-Policy
class site 3
shape average 95000000
  queue-limit 1024 packets
   service-policy ASE-QoS-Policy
...
multiple other sites
...
policy-map ASE-QoS-Policy
 class VoIP
  priority percent 40
  queue-limit 1024 packets
 class Citrix
  bandwidth percent 40
  queue-limit 1024 packets
 class class-default
  queue-limit 1024 packets

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Logically, what you're doing makes sense, but only Cisco "knows" how their queuing really works.  I've long suspected their embedded CBWFQ shapers have their own queues, which you may not have direct control over for setting their queue depths.

Yes, logically, there's a relationship between bandwidth (including shaping) and queue limits, but it depends on multiple factors.  Shaping is also a special case, because although you're trying to emulate a certain link bandwidth, it's not the same.

[mlund]

Hi

On many platforms the Layer2 overhead is not taken into consideration when router calculates the bandwidth. So 5% left for the L2 overhead may not be enough if there is a lot of small size packets.

If this is the case, try to decrease the shape vlaues to maybe 90% and see if the drops is decreasing.

/Mikael