ping size and mtu size

Hello,
I set up a gateway machine with Debian woody an DSL.
Now I find that ping doesn't work with sizes greater than 1424.
After looking in this newsgroup I examined some settings of
MTU size: my /etc/peers/provider includes the following lines:

.. . .
pty "/usr/sbin/pppoe -I eth1 -T 240 -m 1412
.. . .
mtu 1452
.. . .

Modifying the "mtu 1452" line affects the possible ping size,
but it is always limited to about 20...30 bytes lower than the
mtu value.

What to do for sending greater pings i.e. breaking / reassembling
greater packets into pieces?

Thanks in advance,
Uwe

u.dahms <(E-Mail Removed)> wrote:

....

> Modifying the "mtu 1452" line affects the possible ping size,
> but it is always limited to about 20...30 bytes lower than the
> mtu value.

Yep, you thought about the IP header (20 byte)?

> What to do for sending greater pings i.e. breaking / reassembling
> greater packets into pieces?

man ping
/-M

--
Michael Heiming

Remove +SIGNS and www. if you expect an answer, sorry for
inconvenience, but I get tons of SPAM

On Sun, 21 Sep 2003 09:06:29 +0200, u.dahms <(E-Mail Removed)> wrote:
> Hello,
> I set up a gateway machine with Debian woody an DSL.
> Now I find that ping doesn't work with sizes greater than 1424.
> After looking in this newsgroup I examined some settings of
> MTU size: my /etc/peers/provider includes the following lines:
>
> . . .
> pty "/usr/sbin/pppoe -I eth1 -T 240 -m 1412
> . . .
> mtu 1452
> . . .
>
> Modifying the "mtu 1452" line affects the possible ping size,
> but it is always limited to about 20...30 bytes lower than the
> mtu value.
>
> What to do for sending greater pings i.e. breaking / reassembling
> greater packets into pieces?

man ping, and see -M option.

Why are you setting a lower than standard mtu for pppoe? Yes I have read
those ATM theories, but in actual speed tests, default mtu 1492
consistantly gives me better speeds (645/136) than anything in the 1450's
(630/133 @ 1452).

Max -s for ping (when using '-M do' switch) is 28 bytes less than lowest
mtu in the route. So if you are setting mtu 1452, your max -s would be
1452 - 28 = 1424 (which you verified).

--
David Efflandt - All spam ignored http://www.de-srv.com/
http://www.autox.chicago.il.us/ http://www.berniesfloral.net/
http://cgi-help.virtualave.net/ http://hammer.prohosting.com/~cgi-wiz/

David Efflandt wrote:

> On Sun, 21 Sep 2003 09:06:29 +0200, u.dahms <(E-Mail Removed)> wrote:
> > Hello,
> > I set up a gateway machine with Debian woody an DSL.
> > Now I find that ping doesn't work with sizes greater than 1424.
> > After looking in this newsgroup I examined some settings of
> > MTU size: my /etc/peers/provider includes the following lines:
> >
> > . . .
> > pty "/usr/sbin/pppoe -I eth1 -T 240 -m 1412
> > . . .
> > mtu 1452
> > . . .
> >
> > Modifying the "mtu 1452" line affects the possible ping size,
> > but it is always limited to about 20...30 bytes lower than the
> > mtu value.
> >
> > What to do for sending greater pings i.e. breaking / reassembling
> > greater packets into pieces?
>
> man ping, and see -M option.

I don't see that option - So I got ping.shar from the ping page via
www.gnu.org and still couldn't see it in the man, couldn't compile it
either.

Am I looking in the wrong place for a newer ping program ?

>
> Why are you setting a lower than standard mtu for pppoe? Yes I have read
> those ATM theories, but in actual speed tests, default mtu 1492
> consistantly gives me better speeds (645/136) than anything in the 1450's
> (630/133 @ 1452).

I've recently been playing with mtu and pppoatm - for me 1478 is better
than 1500 - pppoatm seems to adds 10 bytes per packet.

I get an atm cell count from my modem, and by experementation can see that
1478(+10) fills 31 cells 1500(+10) fills 31 and 22bytes of the 32nd -
pppoatm pads to fill the 32nd cell - wasting 26 bytes/packet.

It looks like pppoe adds 8 per packet 31 48 byte cells are 1488 so try mtu
of 1480.

Each cell then gets another 5 bytes added AIUI.

Never having the chance to play with pppoe the above may be totally wrong

But the pppoatm mtu of 1478 that I use in the UK does show better than 1500
on speed tests.

Andy.

>
> Max -s for ping (when using '-M do' switch) is 28 bytes less than lowest
> mtu in the route. So if you are setting mtu 1452, your max -s would be
> 1452 - 28 = 1424 (which you verified).
>

Andy Furniss wrote:

> David Efflandt wrote:
>
> > Why are you setting a lower than standard mtu for pppoe? Yes I have
> > read those ATM theories, but in actual speed tests, default mtu 1492
> > consistantly gives me better speeds (645/136) than anything in the
> > 1450's (630/133 @ 1452).
>
> I've recently been playing with mtu and pppoatm - for me 1478 is better
> than 1500 - pppoatm seems to adds 10 bytes per packet.
>
> I get an atm cell count from my modem, and by experementation can see that
> 1478(+10) fills 31 cells 1500(+10) fills 31 and 22bytes of the 32nd -
> pppoatm pads to fill the 32nd cell - wasting 26 bytes/packet.
>
> It looks like pppoe adds 8 per packet 31 48 byte cells are 1488 so try mtu
> of 1480.
>
> Each cell then gets another 5 bytes added AIUI.
>
> Never having the chance to play with pppoe the above may be totally wrong
>

Now I've looked at pppoe I can see I was totally wrong - there is another
22ish bytes to add to that for MAC/other header - which I means that 1458
may be the optimal - I don't have much faith in that figure without being
able to test with a modem which gives a cell count.

It does, however, mean that tweaking to 1450s over 1492 is not going to
gain as much as tweaking from 1500 to 1478 does for me in the UK.

1492 doesn't leave as much wasted space in the last cell with pppoe as 1500
with pppoa and using smaller MTU reduces efficiency per byte because of the
fixed overheads.

Andy.