Policy-map Counter

[AlanGregory]

A simple doubt. Based on the follow config, why packets are dropped before reaching 40 mbps ?

policy-map 35mbps
 class class-default
  police rate 40 mbps
   exceed-action drop
  !
 !
 end-policy-map
!

RP/0/RSP0/CPU0:xxxx1#show qos interface TenGigE0/0/2/1.100.pppoe14948 output
Wed Aug  8 23:04:38.169 GMT
Interface: TenGigE0_0_2_1.100.pppoe14948 output
Bandwidth configured: 10000000 kbps Bandwidth programed: 10000000 kbps
ANCP user configured: 0 kbps ANCP programed in HW: 0 kbps
Port Shaper programed in HW: 0 kbps
Policy: 35mbps Total number of classes: 1
----------------------------------------------------------------------
Level: 0 Policy: 35mbps Class: class-default
QueueID: 65570 (Port Default)
Policer Profile: 62 (Single)
Conform: 40000 kbps (40 mbps) Burst: 500000 bytes (0 Default)
Child Policer Conform: TX
Child Policer Exceed: DROP
Child Policer Violate: DROP
----------------------------------------------------------------------
RP/0/RSP0/CPU0:xxxx#show policy-map interface TenGigE0/0/2/1.100.pppoe14948 output
Wed Aug  8 23:05:09.060 GMT

TenGigE0/0/2/1.100.pppoe14948 output: 35mbps

Class class-default
  Classification statistics          (packets/bytes)     (rate - kbps)
    Matched             :            62626395/85047080230          10688
    Transmitted         : N/A
    Total Dropped       :             2461771/3520342568           250
  Policing statistics                (packets/bytes)     (rate - kbps)
    Policed(conform)    :            60164624/81526737662          10438
    Policed(exceed)     :             2461771/3520342568           250
    Policed(violate)    :                   0/0                    0
    Policed and dropped :             2461771/3520342568

[smilstea]

How QoS works, specifically for a policier, is that when you configure 40mbps its 40mbps across 1s, however we have whats called a queue-limit, and by default that is 100ms, so the 40mbps is broken into segments of 100ms (4mbps segments). If you burst over 4mbps during that 100ms then you will see drops even if during the remaining 900ms you are using less than 36mbps. One way to get around this is to increase the queue-limit, however by doing that you are adding latency.

Thanks,

Sam

[AlanGregory]

Thanks for clarifying.

Does shape have the same behavior ?

[smilstea]

Yes it does.

You can find the current queue-limit, often denoted in kbps or mbps, from show qos interface. The default is 100ms worth of service-rate.

Edit: I should note here that a shaper would be better as it buffers instead of a policier that just drops the traffic. So you would see fewer drops.

Sam

[Aleksandar Vidakovic]

It typically depends on microbursts. You can see that the burst size is set to 500kB. 

There's a section on this in the https://community.cisco.com/t5/service-providers-documents/asr9000-xr-understanding-qos-default-marking-behavior-and/ta-p/3128709 document.

hth,

/Aleksandar