In response to many problems I've had with MPD as a PPTP server for
Windows, as well as other's questions, I set up a test network and
did a bit of experimenting.
In my experiments, I found a number of interesting facts about PPTP:
1) PPTP doesn't do a lot of friendly negiotiation. If the server and
   client aren't configured identically, problems result.  These are
   not always "can't connect" problems, sometimes they are "everything
   works except FTP" problems.
2) disabling non-used network protocols (IPX/SPX & NetBEUI) on Win98
   caused a 50% improvement in speed.
3) The connection is about the same speed whether encryption was turned
   on or off.
4) Disabling multilink allows the PPTP connection to go almost as fast
   as the raw network speed (on Windows XP!)
5) No matter what we did, Windows 98 was always 10x _slower_ than the
   raw network speed.
6) Multilink turned on makes Windows XP as slow as Windows 98
7) I juggled MTUs around until I couldn't stand it any more.  No MTU
   changes that I made cause any noticable change to the connection
   (either speed or reliability)

So my conclusion is this:
1) Use encryption, it doesn't slow anything down.
2) Disable multilink, whatever it does, it's not good
3) Don't use Windows 98, it's a piece of junk
4) Make absolutely sure the client is configured exactly
   like the server.

This is pretty useful information for me, I hope it's helpful for you
as well as the archives.  If this question comes up in the future, point
the questioner to this message in the archives.

[EMAIL PROTECTED] wrote:
On Fri, 31 Jan 2003 10:23:37 -0800 (PST), Bill Moran wrote:
[could you wrap lines around 72 chars or so, please]
Sorry about that.

After connecting via VPN I can get decent throughput from the MPD

host but

very poor speed from anything past it.
What do you mean by this?  We use MPD off and on, and (honestly) it

is just slow.
From other posts I knew MPD would be slow but what concerns me is

that it is how much slower it is beyond the mpd host itself, see test
results below.

I'm not sure I understand your test results.
Are you saying
PPTP client <--> MPD machine <---> "other host"
?
If so, is "other host" on the Internet, or on your
local network?
Other host is on the local network behind the MPD box and transfers
files at a slower rate over the PPtP connection than a transfer from the
MPD box. I also have the 'other host' aliased to a public IP address so
thats how I got the baseline from it.

We've seen that trying to route through the MPD
machine to the internet is terribly slow, but
haven't noticed any problems with routing to the
local network.
Did you check the box on the MS side to say
"use gateway on remote network"?
Nope.

A separte public IP is redirected to a 4.7 RELEASE box on the

inside.
Client(s) tested with have been Windows 2000 SP2 and SP3 from 2

different ADSL Lines.

client-----1.2.3.4 MPD/NATD 172.16.105.80------172.16.105.66 /

5.6.7.8 Redirected from 1.2.3.4
Tests using Penguinet SCP and a 1.9 MB ZIP file. Baseline Download the file from the public IP's 1.2.3.4 -> client 180 kBs 5.6.7.8 -> client 180 kBs
Now test via the PPtP.

172.16.105.80 aka. 1.2.3.4 -> client 84 kBs ****
172.16.105.66 aka. 5.6.7.8 -> client 35 kBs **** These are the results
that don't make sense.

I see now.
We haven't tested this extensively.  We've only seen it when routing
into the VPN, just to
go back out on the Internet (which seemed a silly thing to do).
NO I'm not trying to go back out onto the Internet but could if you
wanted to make sure your remote workers were safe behind your firewall -
but thats a policy/procedure discussion and not for this one :)

Actually I used SCP on the second test so as not to skew things, in
normal operations we won't
> be. My concern is test to 172.16.105.66. What

would make it perform worse than to 172.16.105.80 ? In my mind they
should be same, like the public IP tests.

Apparently, something in MPD isn't working as efficiently as it should.


It also shows Transmit Errors=0 Receive Errors=xx <- increments at a
slow rate when connected.

Ok, now this is something.  We need to find out the nature of the
errors and fix it.

I'm very interested in getting this working better for the same reason
that you are.  I'm going to set up a test network here and see what I
can figure out.  I'll keep in touch with you on my findings if you
agree to do the same.
Certainly, I wonder if Archie Cobbs is out there today :)
Here's a recap,
File downloads to the remote client are much slower from a box(es) on
the same network as the MPD server/Gateway than from the MPD server
itself.
MPD server is also running Natd and IPFW in OPEN mode for this testing.
Have adjusted the MTU down to as low as 1350 with no difference in
performance. ng0 does display an MTU of 1350 when the tunnel is up.
Have tried with compression on/off - no change.
On the W2K Network status I see a steady increase on 'Receive Errors'
when the PPtP is up. Transmit errors=0
Could it be something to do with NATd ? Since I'm already behind on this
by 4 days I think I'll do up a test network without NAT and see.
If someone can read a tcpdump I can do one of those too. Let me know
from which box and what options.

Thanks, John.

__________________________________________________________
Get your FREE personalized e-mail at http://www.canada.com



--
Bill Moran
Potential Technologies
http://www.potentialtech.com


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-questions" in the body of the message

Reply via email to