Cisco Support Community
Community Member

QoS between two FE on 7206

We have build up a test environment, where two FE interfaces (in and out) of a packet generator (smartbits) are connected to two corresponding ifs on the 7206. We have set up CBWFQ, based on DSCP-values generated by the packet generator.

When we use 128-byte packets, we see a drop rate of about 60-100% when bandwidth utilization (generated by smartbits) goes beyond 50%. When we use 256-byte packets or higher the drop rate goes down to 0% and all traffic is forwarded.

With the small packets we can see a very high cpu utilization (up to 95%) where there is only 60% in peak using the bigger packets.

Is this a bug or a feature (means, limitation by hw;-). Thanks for feedback.

Cisco Employee

Re: QoS between two FE on 7206

it's hard to say.

We would need more information to decide.

CPU type, software version, amount of memory, interface module, config, ...

It could be hardware limitation, since all HW is limited by BW and packet/sec.

So, with 128bytes packet, you send 2 times more packets than with 256 bytes packet for the same amount of BW.

Could also be software since there was a lot of issues related to QoS in the path.

Finally, in the real world you have a distribution of size. So, you should not see this problem.

Community Member

Re: QoS between two FE on 7206

Thanks for the feedback. What do you mean with more information? Do i have to open a tac case or can we discuss this in the forum? If forum is o.k. because other people could be interested in the ongoing i would post some configuration infos in the next message.

Community Member

Re: QoS between two FE on 7206


Is it a 7206 VXR? If yes, CEF is enabled by default, but for any reason have you disabled it?


Nuno Morais

Community Member

Re: QoS between two FE on 7206


yes, it is a 7206VXR with a NPE-400 and 256MB of RAM.

No, CEF is not disabled.

We have build up a class-map, which matches the dscp values generated by the packet generator. Those classes are build into an output policy, where we define bandwidth restrictions for the different classes. Also we have defined a 100% max-reserved-bandwidth on the output interface.

So, we use CBWFQ as a queuing mechanism - and the 'sh policy-map'-command tells us, everything is fine configured.

As i have said, there is no problem (means, no drops) when we use >=256byte packets. When we use packets <=128byte there are drop rates at about 70-80%!


CreatePlease to create content