lkml.org 
[lkml]   [1996]   [Mar]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Slow PPP?
On Fri, 29 Mar 1996, Mike Kilburn wrote:

> > 100Kb test file Winsock ppp-2.2.0e
> > -----------------------------------------------------------------
> > /pub/test/emptyfile (compressible) 6.24 Kb/s 3.80 Kb/s (-40%)
> > /pub/test/fullfile (non-compressible) 3.11 Kb/s 2.80 Kb/s (-10%)
> >
> > I use the same box booted into Dos or Linux, so there are no hardware
> > changes, and the transfers were done <10 minutes apart to avoid changes at
> > the ISP's end where possible.
> >
>
> How did you do the timing? You should time from the first byte to the last
> byte received *on the receive side*. Buffering differences can lead to vastly
> different results when viewed from the transmit side, and unless your using
> the *exact* same program (not possible if its with 2 OS ) you can get
> different results. The fact that you got exactly 3.80 Kb/s on a highly compress
> able file with a link of 28.8 tells me there could be a port fixed at 38.4
> somewhere in the stream. Here are some other things to consider:
>
Yes I was measuring at the receive end. My ISP has provided the files I
mentioned for just this type of testing. Of course I have to trust ftp to
report correct throughput

> 1. Can you be sure that you ISP has *all* their TS ports at 115200? Maybe
> you got a 38.4 port when you dialed with linux.

They claim to run 115,200 everywhere, and I've never seen the slowdown
associated with a 38.4 port before... I can only trust them :-)

> 2. Can you be sure the Linux box is at 115200?

I've already made a note to do a hard-and fast check on this over the
weekend ! I'll be asking the modem to report DTE and link speeds during
the dial process. I did this when it was first set-up, but perhaps I've
changed somethig recently which is causing the problem... Wouldn't it be
nice if it were that simple.

> 3. Can you verify the asyncmap used when dialing with Linux is the
> same as Windows. This can make a big difference when comparing
> thruput.

OK, how do I discover/set the asyncmap under linux ? Would it really cause
THAT much difference ? I know what I'm using under DOS so I could
try to duplicate it to remove the variable

> 4. Can you check if options like V.J header compression are the same.

V.J header comp, mtu, mru are all reported the same.

I'll get back to the group on Monday then... Or perhaps I'll wait 'til
Wednesday when I'm in Amsterdam... Boy do they sell good coffee! :-)

Steve



\
 
 \ /
  Last update: 2005-03-22 13:36    [W:0.714 / U:0.016 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site