To summarize my comments:
The benefit of using CBWFQ (class based weighted fair queuing) with the fair-queue statement in the class-default queue (your second example):
...
class class-default
fair-queue
queue-limit 1024 packets
...
- Packets use WFQ (weighted fair queuing) for scheduling and are de-queued based on their calculated weight.
When using CBWFQ without the fair-queue statement in the class-default queue (your first example):
...
class class-default
bandwidth remaining percent 20
random-detect dscp-based
...
- Packets are de-queued using the FIFO (first in first out) method.
1) Am I correct that "30 second offered rate" is a bandwidth which comes into service-policy named traffic-policer?
30 second offered rate is based on load-interval command, you can change it under interface configuration. Default is 5 minutes.
Router(config)# interface FastEthernet 0
Router(config-if)# load-interval 30
It is a length of time for which are data used to compute load statistics. So simply said, you can imagine it as the average traffic for 30 sec interval.
And the all traffic which is processed by policy-map traffic-policer is specify above that line in class class-default (545169 packets, 822918951 bytes)
2) Why there are two lines "shape (average) cir" and "target shape rate"?
This is really good question. Many people are confused, I think that it has something in common that Cisco does not have a good explanation about it in any book.
You will never see different values between Shape average and target shape as long as you use shape average command. In case you will configure your shaping map by shape peak command instead of shape average your shaper will starts to use PIR instead of CIR.
What PIR means? How will change it your shaping? That's quite long story. I am apologize but I am not going into any explanations but you can find answer to your questions here:
http://blog.ine.com/2008/08/26/understanding-the-shape-peak-command/
3) Am I correct that this part:
queue stats for all priority classes:
queue limit 64 packets (queue depth/total drops/no-buffer drops)
0/0/0 (pkts output/bytes output) 190/10640
..is only for GRE-traffic traffic-class as it is the only one with
priority configuration?
These are statistics for priority low latency queues. And you're right. As long as you use only one class (GRE-traffic) configured as priority, statistics are recorded only from mentioned traffic.
4) Am I correct that this 19.5Mbps(shape average 19500000) is shared
between all the traffic classes specified in child service-policies?
class-default and GRE-traffic in my case.
You're indeed right, CIR 19.5Mbps is shared by all child classes that you're configured in policy-map prioritize-GRE-traffic.
Best Answer
Have you tried using
bandwidth qos-reference?