Plusnet Slow web browsing MTU change!

Hoping someone can help here.
Been through the mill with Plusnet support.
Plusnet support looking to get BT to check my line which may cost me 46ukp +
vat.
Problem:
Really really slow connections to web sites especially USA web sites like
msnbc.com or ebay.co.uk.

MTU and RWin were set to default initially.

Since changing MTU to 576 my web browsing speed has increased immensely.
RWIN is at 55600.

But Plusnet and other websites that do mtu rwin tests are stating that a
higher setting would be better. Problem is that a higher setting completey
wrecks my browsing. I have to click on the link and then a split second
later click on the link again for the web page to display if it displays at
all, other pages just time out.

I am going through a wireless browser and anytime I change MTU setting I do
it at the router and at my PC using DRTCP.

I did an etherreal trace when my MTU was higher and got messages saying
things like fragmented packet, this is the initial reason I dropped my MTU
to 576, to try and stop these fragmented errors.


Has anyone got any ideas.

In article <433d77fb$0$49777$(E-Mail Removed)>,
(E-Mail Removed) says...

<snip>
> Has anyone got any ideas.

Firstly, stop plusnet from involving BT. This is 99% likely a problem
with your setup and you probably don't want to be potentially charged
for BT investigating.

Secondly, what results do you get if you set the MTU on the router to
1500 and your PC to 1430 (or less)? You should not change the MTU on the
router to anything lower than 1500 - do any necessary changes on the PC.

Have you played around with anything else on the router, e.g MRU or
other settings? Do you get the same problems when wired in to the router
as opposed to wireless?

Have you tried fully resetting the router to firmware defaults *and*
reinstalling the network/wireless cards from device manager (remove and
let windows redetect them - they'll get reinstalled with the default
settings)? Try updating your wireless drivers if you only experience the
fault on wireless but not the wired connection. Update the router
firmware if a later version is available from the manufacturer.

Let us know how you get on,

Alex

On Fri, 30 Sep 2005 18:36:41 +0100, "mmmmm" <(E-Mail Removed)>
wrote:

>Since changing MTU to 576 my web browsing speed has increased immensely.
>RWIN is at 55600.


a bigger RWIN may help. By making MTU smaller you make the RWIN equal
to more packets = more time, ideal for US traffic. Doubling RWIN or
more could have the same effect - try 250,000

Phil
--
Tiscali - dialup speeds at Broadband prices, see
http://bbs.adslguide.org.uk/postlist...&Board=tiscali

AOL - the unlimited ISP of choice for heavy downloaders.

On Fri, 30 Sep 2005 18:36:41 +0100, mmmmm
<(E-Mail Removed)> wrote:

> I did an etherreal trace when my MTU was higher and got messages saying
> things like fragmented packet, this is the initial reason I dropped my MTU
> to 576, to try and stop these fragmented errors.

What were the messages "saying things like fragmented packet"? Were
they messages relating to TCP segments or were they ICMP messages?

And what made you think they were "errors"? Fragmentation itself is
just a feature of IP and as such fragmentation isn't an error, but
errors relating to fragmentation can occur if things like the
reassembly timeout is set too low, or if you have the DF flag set on
outbound packets and don't allow inbound ICMP itype 3 code 4 messages.
You have not really provided enough information for anyone to even
start to guess what the problem may be.

> Has anyone got any ideas.

You seem to have a totally screwed up setup since even if
fragmentation is necessary, you should still be able to transfer data,
albeit somewhat less efficiently then would otherwise be the case.
Have another look your Ethereal capture logs since they should give
you a good idea what, and maybe where, the problem lies.

On Fri, 30 Sep 2005 19:00:52 +0100, Alex Crosby <(E-Mail Removed)>
wrote:

>Secondly, what results do you get if you set the MTU on the router to
>1500 and your PC to 1430 (or less)? You should not change the MTU on the
>router to anything lower than 1500 - do any necessary changes on the PC.

Could you please explain this more? I have always read that MTU should
be the same on the router and PCs.

TIA

--
Cheers,

Guy

** Stress - the condition brought about by having to
** resist the temptation to beat the living daylights
** out of someone who richly deserves it.

In article <(E-Mail Removed)>,
(E-Mail Removed) says...
> On Fri, 30 Sep 2005 19:00:52 +0100, Alex Crosby <(E-Mail Removed)>
> wrote:
>
> >Secondly, what results do you get if you set the MTU on the router to
> >1500 and your PC to 1430 (or less)? You should not change the MTU on the
> >router to anything lower than 1500 - do any necessary changes on the PC.
>
> Could you please explain this more? I have always read that MTU should
> be the same on the router and PCs.

It's not necessarily a bad thing that they're both the same, however, if
you've got other PCs on the network that have the windows default MTU
set the router will have to fragment those packets as its MTU is lower
than on the client PCs. With MTU at 1500 on the router, you're safe
either way - the router won't fragment the packets and you'll get the
benefit of the specific sub-1500-MTU you set on your PC (the router
won't touch the packets other than translating the address as
necessary).

HTH
--
Alex Crosby

"is the search function a database or is it people seeing i=3Fm searching
and helping out?"
-Anon

On Sat, 1 Oct 2005 04:12:29 +0100, Alex Crosby
<(E-Mail Removed)> wrote:

>It's not necessarily a bad thing that they're both the same, however, if
>you've got other PCs on the network that have the windows default MTU
>set the router will have to fragment those packets as its MTU is lower
>than on the client PCs.

It's okay, I've changed the MTU on all PC's on the LAN.

I'll save myself some work next time and do it 'your' way <g>.

Thanks.

--
Cheers,

Guy

** Stress - the condition brought about by having to
** resist the temptation to beat the living daylights
** out of someone who richly deserves it.

Wondering why you need to play around with MTUs, I'm on PlusNet and I've
been using the default setup on my DG834GT and the rest of the PCs (2
Linux machines and 2 Windows).. I've no problem with the speed on any of
them.

Plusnet is not even sending my replies to this group. Gawd I hate
Plusnet!

Anyway back to the issue at hand.

My router will not go above 1492 in manual mode. I type 1500 and it did
not like it!

The etherreal fragmentation message is an ICMP message and it seems to
have a send address of my router and a dest address of my PC! Here's a
bit of the capture.

"11", "0.741116", "192.168.1.65", "signin.ebay.co.uk", "SSLv3",
"Application Data, [Unreassembled Packet]"
"12", "0.741466", "192.168.1.65", "signin.ebay.co.uk", "SSLv3",
"Continuation Data, [Unreassembled Packet]"
"13", "0.759121", "192.168.1.1", "192.168.1.65", "ICMP", "Destination
unreachable (Fragmentation needed)"
"14", "0.936166", "signin.ebay.co.uk", "192.168.1.65", "TCP", "[TCP Dup
ACK 9#1] https > 3817 [ACK] Seq=1140 Ack=283 Win=17040 Len=0
TSV=52664172 TSER=47727"

So if I reduce my MTU on the home computers to below 1492 but leave the
router on Auto or even manual @ 1492, should I see a speed increase?

Thanks.



WiraOne wrote:
> Wondering why you need to play around with MTUs, I'm on PlusNet and I've
> been using the default setup on my DG834GT and the rest of the PCs (2
> Linux machines and 2 Windows).. I've no problem with the speed on any of
> them.

On Sat, 1 Oct 2005 04:12:29 +0100, Alex Crosby
<(E-Mail Removed)> wrote:

> It's not necessarily a bad thing that they're both the same, however, if
> you've got other PCs on the network that have the windows default MTU
> set the router will have to fragment those packets as its MTU is lower
> than on the client PCs. With MTU at 1500 on the router, you're safe
> either way - the router won't fragment the packets and you'll get the
> benefit of the specific sub-1500-MTU you set on your PC (the router
> won't touch the packets other than translating the address as
> necessary).

For most people the most significant impact of setting a lower MTU is
that TCP connections will be negotiated with a lower MSS, thereby
limiting the size of inbound segments. If your router supports MSS
clamping then that is a far better way of achieving the same effect
since you do not have to change any MTUs (except maybe the MTU for
the PPPoA interface on the router, but only in cases where the router
can only be configured to clamp the MSS relative to an MTU, rather
than clamp it to a value of your choosing).

For example, all the Ethernet MTUs on my local network are set at 1500
bytes, but any TCP connections to/from the Internet are 'clamped' (by
the router) to have an MSS of 1438 bytes, giving an 'effective' MTU of
1478 bytes for the PPPoA connection; 1478 bytes being the value that
should give maximum throughput given the AAL5 overheads and fixed size
of ATM cells. But should a packet of 1500 bytes be sent or received
by the router, it will still be forwarded without fragmentation.