[c-nsp] Traffic Counters and QoS Policy

[Oliver Boehmer (oboehmer)]

Ivan <> wrote on Tuesday, April 01, 2008 11:51 AM:

> I am sending some test traffic through a Cisco 1841 (12.4(15)T4) and
> am hoping someone can confirm what I have found.
> 
> I am sending 64 byte packets (including IP headers) through uniformly
> at 110000bps (215 PPS).  The counters on the interface show traffic of
> 134000bps and 215 PPS.  This works out to be 78 bytes per packet.  It
> would seem the additional 14 bytes are the layer 2 Ethernet header (6
> bytes source MAC, 6 bytes dest MAC, 2 bytes ether type).
> 
> 
> The documentation from Cisco
>
(http://www.cisco.com/en/US/docs/ios/12_0/interface/command/reference/ir
showin.html#wp1017950)
> states
> 
> Five minute input rate,Five minute output rate = Average number of
> bits and packets transmitted per second in the last 5 minutes.
> bytes input = Total number of bytes, including data and MAC
> encapsulation, in the error free packets received by the system.
> 
> 
> The description of the input rate doesn't really clarify if it
> includes the layer 2 header but the byte count does.  Resetting the
> counters and doing the maths shows to me the the input rate counters
> do include layer 2 headers.  Can anyone confirm I am on the right
> track here? 

yes, you are.. interface counters (including the input/output rate which
is calculated based on these counters) include L2 encaps overhead.

> The interesting part is that when I have a QoS policy on the interface
> the policy counters are also using the layer 2 headers in the
> calculations as I get a "30 second offered rate 134000 bps".  Does
> this mean for example if I apply some kind of QoS policy to 5Mbps
> (5000000bps) it will be including the layer 2 headers in the
> calculations?  I had always assumed the it was only layer 3 and above.

same as above: QoS generally includes L2 overhead (it better does as
we're reserving link bandwidth, and L2 overhead can make quite a
difference on the link)..

	oli