Herarchical QoS with class based shaping

[Dariusz Lojewski]

Hi everybody !

As far as I know the purpose of combining the two, shaping and CBWFQ, is because the latter requires congestion to engage, the former can create congestion sooner than the physical interface.

My question is, when the congestion from the shaper takes place and the scheduler of the CBWFQ begins to fill appropriate queues?

Is it when the transmission speed > CIR?

Or

When the shaping buffer(queue) overflows?

By the way, what is the shaping queue (buffer) in this case? Is it a “standalone” queue or it simply consists of the class-based queues ?

By the interrogation of the Policy-Map on the wan-interface we can see, that a queue for the class-default (Shaper) was created with the limit 16250 packets? What is the purpose of the queue? I think , it has nothing to do with shaping-queue.

Extraction from sh policy-map int Gi2/0/0:

Service-policy output: shape-dxt<->dxtt

  Counters last updated 00:00:03 ago

    Class-map: class-default (match-any)

      153103779 packets, 16570964644 bytes

      30 second offered rate 54000 bps, drop rate 0 bps

      Match: any

      Queueing

      queue limit 16250 packets

      (queue depth/total drops/no-buffer drops) 0/0/0

      (pkts output/bytes output) 153103522/16570934386

      shape (average) cir 65000000, bc 260000, be 260000

      target shape rate 65000000

Extraction from the running-file:

policy-map qos-dxt<->dxtt

  class mgmt-cs0-2

    bandwidth percent 5

    random-detect

    random-detect precedence 0 240 480

  class as-nem-cs1-afx

    bandwidth percent 16

    random-detect

    random-detect precedence 1 500 1000

  class gn-gi-af2x

    bandwidth percent 26

    queue-limit 1200 packets

  class ls2-af41

    bandwidth percent 14

    queue-limit 720 packets

  class ls3-af42-cs4

    bandwidth percent 15

    queue-limit 920 packets

  class mgmt-cs2

    bandwidth percent 5

    queue-limit 320 packets

  class routing-cs6

    bandwidth percent 5

    queue-limit 320 packets

policy-map shape-dxt<->dxtt

  class class-default

    shape average 65000000

   service-policy qos-dxt<->dxtt

interface Gi2/0/0

service policy output shape-dxt<->dxtt

[Akash Agrawal]

Hi,

As per explanation i have read in cisco books, interaction between parent shaper and child CBWFQ works as below

"Shaper decides whether packet needs to be put into shaping queue or not and when to take packet from the shaping queue. But between this, queuing scheduler comes into picture and decides which packet needs to service next from the shaping queue."

So each class queue of child policy-map will be a shaping queue. When routing decision confirms that packet will go out of the interface where service-policy is applied, it first checks whether shaping is active or not, Shaping becomes active when single packet exceeds the traffic contract. If shaping is not active traffic can directly be put to hw tx ring.

Once shaping is active and packets are getting queued, shaper will when to dequeue one packet based on rate configured in the shaper but which packet to be dequeued will be decided by Scheduler as per preference given in child policy-map.

Here i think, queue length will be limited for each class by queue-limit of that class or WRED max threshold (WRED max threshold can not exceed queue-limit of the class) and shaper queue length is more than sum of queue-limit of each child class, it will not play any role.

I just tried to configure a parent-child service-policy and buffer-limit on shaper is calculated as sum of max buffer defined in the child policy-map

R21_7206#show policy-map parent
  Policy Map parent
    Class class-default
      Traffic Shaping
         Average Rate Traffic Shaping
         CIR 65000000 (bps) Max. Buffers Limit 1000 (Packets) <<<<<<<<<<
      Bandwidth 200 (kbps) Max Threshold 800 (packets)
      service-policy child
R21_7206#
R21_7206#sh policy-map child
  Policy Map child
    Class EF
      Bandwidth 50 (%) Max Threshold 500 (packets) <<<
    Class AF
      Bandwidth 50 (%) Max Threshold 500 (packets) <<<<
R21_7206#

-- Pls dont forget to rate the helpful posts --

Regards,

Akash

[Dariusz Lojewski]

Hi guys, 

thanks for your responds and help. They acknowledged what I had assumed, so now I can sleep well :)

Best regards.

[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

My question is, when the congestion from the shaper takes place and the scheduler of the CBWFQ begins to fill appropriate queues?

Is it when the transmission speed > CIR?

Or

When the shaping buffer(queue) overflows?

By the way, what is the shaping queue (buffer) in this case? Is it a “standalone” queue or it simply consists of the class-based queues ?

In theory, shaped queuing should work with directly with class queues, but from my experience, I suspect, at least on some platforms and IOS versions, that there might be a shaper queue too.  I've haven't seen/found public Cisco documentation for actual implementation details.