EF traffic getting remarked, but traffic never exceeds police CIR value

[Robert Burlingame]

Hello, so I'm trying to police voice (EF) traffic outbound from the router and limit it to 1.536Mbps.

We're doing this to match our carrier's CAR profile. However, when I show the policy-map stats, it looks as if my exceed action happens before the traffic volume gets to 1.536Mbps.

Config:

interface GigabitEthernet0/1
 description MPLS
 ip address 7 255.255.255.255
 service-policy output MyPolicyMap

!

class-map match-any VOICE
 match ip dscp ef 
!
policy-map MyPolicyMap
 class VOICE
  police 1536000 19200 conform-action transmit  exceed-action set-dscp-transmit af41

RTR-1#sh policy-map interface g0/1
 GigabitEthernet0/1 

  Service-policy output: MyPolicyMap

    Class-map: VOICE (match-any)  
      750486 packets, 235065040 bytes
      5 minute offered rate 64000 bps, drop rate 0000 bps
      Match: ip dscp ef (46)
      police:
          cir 1536000 bps, bc 19200 bytes
        conformed 748830 packets, 233225375 bytes; actions:
          transmit 
        exceeded 1656 packets, 1839665 bytes; actions:
          set-dscp-transmit af41
        conformed 58000 bps, exceeded 1000 bps

I can run this command every 5 seconds and the exceeded counter continues to climb while the 5 minute offered rate never reaches the cir of 1536000. What am I missing here? How can I see the true EF rate?

(I've even set the "load-interval" to 30 seconds. Never hits 1536000)

Thanks,

Robert

[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

The policer is marking overrate traffic based on its Tc value, which is usually in the millisecond range (usually something like 4 to 25 ms).

[Robert Burlingame]

Joseph, thanks for the comment. Can you please explain your comment in relation to the show command I've posted?

[Joseph W. Doherty · Accepted solution]

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

cir 1536000 bps, bc 19200 bytes

Bc/CIR = Tc

19200 bytes * 8 bytes/bit  / 1536000 bps = .1 sec or 100 ms

Not within the example subsecond range I noted, but still subsecond.

Actual traffic is always at line rate.  The policer "counts" amount of traffic within Tc, and drops excess with that Tc.

[Robert Burlingame]

Thanks Joseph!

I bumped the burst value up to 192000. That made our interval 125ms. Since then, we're not seeing too many marked packets.

[trfinkenstadt]

Robert,

You might also consider doing "load-interval 30" on your interface which will give you 30sec instead of 5 minute average bandwidths.  Still probably not granular enough to see if you actually exceed your rates but it'll look a little better.

..

[Robert Burlingame]

trfinkenstadt, thanks for the suggestion. Not sure how much that changes the results, but, it does provide a more frequent metric and a little better graph.

[Joseph W. Doherty]

BTW, I figure your Tc is now 1 second, a 10x increase.  (NB: That's not to say its bad doing this.  However, you did say your were trying to match your carrier.  If you burst higher then the carrier's measurement allows, they might drop.)

[Vinit Jain]

What is the Platform and the software version you are using?

[Robert Burlingame]

4331 running IOS-XE

sr4300-universalk9.03.15.01.S.155-2.S1-std.SPA.bin