[c-nsp] Cisco ASR Shaping (Very Basic QoS) - Output Drops

[Troy Boutso]

Hey

I look after global MPLS network which consists of multiple sub-line rate -
Layer 2 circuits. Meaning I usually get a 1Gbps ethernet hand-off. But the
services are usually configured by the provider as 200Mbps or 300Mbps point
to point circuits. All my circuits terminate on Cisco ASR1000 Series
routers. Specifically the Cisco ASR1001X routers.

Currently running: Version 03.15.02.S - Standard Support Release


In terms of capacity we are usually pretty good. We never tend to go over
50% utilization (average), though of coarse there are always bursty periods
which could use all available bandwidth for very short periods of time.

What I have done in the past (whilst working for smaller service providers)
is to apply a very basic Shaping policy on all these sub-line rate links to
"smooth" out any bursty traffic. I've never really given this a second
thought. It seems to work fine.

However lately I've been noticing extended periods where the drop rate is
rather high. Usually I notice it for a few seconds. Then it stops. But I'm
noticing it alot more and it is failry constant.

The overall traffic demand hasn't increased, so not sure what could be
causing this all of a sudden. Perhaps a new application on the network is
more "Bursty".


Here is the output of show policy-map int (sometimes the drop rate is
1Mbps) - That is high!

*sh policy-map int Gi0/0/1 *
* GigabitEthernet0/0/1 *

*  Service-policy output: SHAPE198MBPS*

*    Class-map: class-default (match-any)  *
*      226356713752 packets, 130403313955357 bytes*
*      30 second offered rate 101473000 bps, drop rate 88000 bps*
*      Match: any *
*      Queueing*
*      queue limit 512 packets*
*      (queue depth/total drops/no-buffer drops) 0/43866538/0*
*      (pkts output/bytes output) 226208256438/130353031064610*
*      shape (average) cir 198000000, bc 792000, be 792000*
*      target shape rate 198000000*



Here is the basic config. Yes very basic.

*!*
*policy-map SHAPE198MBPS*
* class class-default*
*  shape average 198000000   *
*  queue-limit 512 packets*
*!*
*policy-map SHAPE290MBPS*
* class class-default*
*  shape average 290000000   *
*  queue-limit 512 packets*
*!*


*!*
*interface gi0/0/1*
* service-policy output SHAPE198MBPS*
*!*




Is there anything I can do to make this more effective? Am I overlooking
something simple?
I do not want to implement anything too elaborate. My goal is to not
calssify all the different traffic types in the network and allocate BW. I
can address performance elsewhere.

I want to make sure I shape the entire link to the desired rate with
minimal drops. Should I increase something?

Thanks you very much in advance.

Troy