[c-nsp] ASR920 Microbursts

Tue Aug 11 03:38:01 EDT 2015

Dear George,

The drops that I saw was due to the tail drop of the queue. I understood that the drops was caused by a burst during a brief space of time. I saw the queue depth with packets before I saw the drop counter incremented. I think that the issue are the microbursts, but how can I see the counter for the encapsulation mistmach? Do you have information?

Thanks.

ASR920-ACCR01-01#sh policy-map int g0/0/21 ser ins 10 out

GigabitEthernet0/0/21: EFP 10

  Service-policy output: POLICY_CUSTOMER_EGRESS_33Mbps

    Class-map: CUSTOMER_EGRESS_PREC_ALL (match-any)

      31429576 packets, 6840985324 bytes

      30 second offered rate 7000 bps, drop rate 0000 bps

      Match:  precedence 0  1  2  3

      Match:  precedence 4  5  6  7

      Queueing

      queue limit 15000 us/ 61875 bytes

      (queue depth/total drops/no-buffer drops) 0/2081801/0

      (pkts output/bytes output) 29347775/6680683884

      shape (average) cir 33000000, bc 132000, be 132000

      target shape rate 33000000

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

      5 packets, 324 bytes

      30 second offered rate 0000 bps, drop rate 0000 bps

      Match: any

      queue limit 393 us/ 49152 bytes

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

      (pkts output/bytes output) 5/324

De: George Giannousopoulos [mailto:ggiannou at gmail.com] 
Enviado el: lunes, 10 de agosto de 2015 14:06
Para: Jordi Magrané Roig <jordimagrane at hotmail.com>
CC: cisco-nsp <cisco-nsp at puck.nether.net>
Asunto: Re: [c-nsp] ASR920 Microbursts

Hi,

Are you sure the drops are due to microbursts?

We have an ongoing issue with drops on ASR920 and TAC informed us that the same counter is used for the mismatched encapsulation packets too.

That means, no matter how big your buffer is, you may still see some drops on your interface, depending on your configuration.. 

--

George

On Tue, Aug 4, 2015 at 5:46 PM, Jordi Magrané Roig <jordimagrane at hotmail.com <mailto:jordimagrane at hotmail.com> > wrote:

Dear Colleagues,

Recently I have installed one ASR920 and I have configured on 1G interface
one service instance with an outbound policy-map shaping to 30 Mbps. The
problem is that I noticed that the ASR920 has the same microburst issue than
the ME3600. I have tried to adjust the queue-limit in order to avoid drops
but then the latency increments. I have configured:

policy-map POLICY_CUSTOMER_EGRESS_30Mbps

class class-default

  shape average 30000000

I have tried also with different classes of service but the issue still
persist.

I would like to know the recommendation about the queue-limit size, the
relation with the values of burst committed and then what should be the
recommended configuration of shaping, recommendations about fine tunning the
shaping. I have noticed that there Cisco platforms that shape better, for
example the ASR9001 or 7600 with ES+ shape better than ASR1000 with
SPA-5X1GE-V2 or ME3600.

Another problem is that I didn’t find enough information about QoS on ASR920
platform. Somebody knows the default queue size of 1G port? How the packet
buffer works? Is it like ME3600 (shared queue buffer per ASIC)?

Thanks!

_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net <mailto:cisco-nsp at puck.nether.net> 
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/