[c-nsp] mpls vpws (p-to-p pw) mtu understanding

Aaron aaron1 at gvtc.com
Tue Apr 24 10:30:26 EDT 2012


Folks, please see my notes, .yesterday I tested veex-injected traffic frame
rate of 9191 (won't work) then 9190 worked.  See the following settings I
had in my network when 9190 worked.  Please help me understand what you
think the 16 bytes might be regarding the observed frame rate from VeEX
tester of 9190 and what I believe is the underlying service interface of the
PW..  lemme know if I'm making an incorrect comparison with comparing the
VeEX 9190 to the PW MTU of 9206.. I wonder if I should be comparing veex
frame size of 9190 to the gige phy interfaces of the end point 901's which
are 9216. ?

 

I have.  (asr901's are actually the pe's..9k's are p's)

 

veex-----g0/0 (asr901) g0/4------g0/0/0/0 (asr9k)-------mpls
cloud------(asr9k)g0/1/0/1------g0/4(asr901)g0/0------veex

 

-----------------my notes from testing on
4/23/2012--------------------------------------------------------

 

VeEX Tester.... 9190 frame size on VeEX tester worked with the following mtu
settings on 901 and 9k...

 

--- asr901 gige interfaces 9216 (default)

--- asr901 vlan 1 set to 9202* (defaults to 1500)

--- asr9k tengige and gige ints set to 9216 (i think default is 1514**)

--- asr9k pw mtu 9206 (default)

--- asr901 pw mtu 9206 (defaults to 9216)

 

* this matches (-14 which is cosmetic in asr9k as it renders frame sizes
WITH eth hdr in show outputs) asr9k ip interface of g0/0/0/0 (ip shared
segment with asr901).  mtu match on shared ip ints when running ospf is a
must...OR ospf mtu ignore is needed.  i opted for mtu consistency

** again, asr9k renders mtu sizes in show outputs WITH eth header, whereas
it's mtu is really that minus the 14 byte eth hdr)

 

...not sure what the 16 is (9206 pw mtu - 9190 (the mtu (frame size) seen on
veex tester) = 16) ...possibly 4 for the veex 1q tag of vlan 10, then
perhaps another tag inserted by service instance of 901 for another vlan 10
tag (tag rewrite pop 1 symm statement didn't seem to change that as we
tested it) then maybe two mpls labels... 

 

4 - veex 1q tag

4 - 901 1q tag

4 - pw label (i think there is one for pw)

4 - core routing mpls label

= 16

 

...my quesses here based on the aforementioned numbers i'm seeing.

 

-----------------end, my notes from testing on
4/23/2012--------------------------------------------------------

 



More information about the cisco-nsp mailing list