Re: HDLC Overhead, serial interface graphs

Hi Anjum,

The answer to your question appears to be both "yes" and "no" all at
the same time. You can lab this and gain some good insight if you
have Dynamips up and running somewhere. First, take two serially
connected routers (R1 and R3 in the example I'm about to post).
Disable CDP and HDLC keepalives on the serial interfaces so there's no
clutter. Clear the counters on your routers and validate that bytes
in and out remains constant at zero. Start the capture function of
Dynagen on one of the serial interfaces. Do the following:

Rack1R1#ping 10.0.13.3 rep 1 size 100

Type escape sequence to abort.
Sending 1, 100-byte ICMP Echos to 10.0.13.3, timeout is 2 seconds:
!

And then issue:

Rack1R1#sh int s2/1
<snip>
      1 packets input, 104 bytes, 0 no buffer
<snip>
      1 packets output, 104 bytes, 0 underruns

Now let's have a look at the capture file:

No. Time Source Destination
Protocol Info
       1 0.000000 10.0.13.1 10.0.13.3
ICMP Echo (ping) request

Frame 1 (104 bytes on wire, 104 bytes captured)
Cisco HDLC
Internet Protocol, Src: 10.0.13.1 (10.0.13.1), Dst: 10.0.13.3
(10.0.13.3)
     Version: 4
     Header length: 20 bytes
     Differentiated Services Field: 0x00 (DSCP 0x00: Default; ECN: 0x00)
     Total Length: 100
     Identification: 0x0002 (2)
     Flags: 0x00
     Fragment offset: 0
     Time to live: 255
     Protocol: ICMP (0x01)
     Header checksum: 0x8d93 [correct]
     Source: 10.0.13.1 (10.0.13.1)
     Destination: 10.0.13.3 (10.0.13.3)
Internet Control Message Protocol
     Type: 8 (Echo (ping) request)
     Code: 0 ()
     Checksum: 0x8f79 [correct]
     Identifier: 0x0002
     Sequence number: 0 (0x0000)
     Data (72 bytes)

0000 00 00 00 00 00 06 ee c8 ab cd ab cd ab cd ab cd ................
0010 ab cd ab cd ab cd ab cd ab cd ab cd ab cd ab cd ................
0020 ab cd ab cd ab cd ab cd ab cd ab cd ab cd ab cd ................
0030 ab cd ab cd ab cd ab cd ab cd ab cd ab cd ab cd ................
0040 ab cd ab cd ab cd ab cd ........

So we have an IP header length of 20, an ICMP header length of 8, and
ICMP data of 72 bytes, which obviously totals the 100 bytes I called
for in my ping command. So there are four bytes that can be
attributed to the HDLC *header*. However, I don't think the story
ends there. Check out:

http://en.wikipedia.org/wiki/Cisco_HDLC

So if the above is to be believed, then, as is often the case, neither
the interface counters nor Wireshark are accounting for/displaying the
FCS+Flag trailer (three bytes total per frame). The interface
counters *do* account for the keepalives that will normally be present
(you can observe this by simply re-enabling keepalives and clearing
counters to have a look). That being the case, I guess the answer is
that the interface counters account for a little better than half of
the total HDLC overhead. Either way it's pretty miniscule in the
grand scheme of things...

Regards,

Scott

On Aug 25, 2009, at 3:48 , Muhammad Anjum wrote:

> Hi All,
>
> If we draw mrtg graphs of a serial interface, will graphs show total
> number of packets in and out, include HDLC header (HDLC overhead).
>
>
> Thanks,
>
> Anjum
>
>
> Blogs and organic groups at http://www.ccie.net
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html

Blogs and organic groups at http://www.ccie.net
Received on Tue Aug 25 2009 - 08:38:48 ART