Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 16 Dec 1998 10:28:02 -0000
From:      "Bond, Jeffery" <Jeff.Bond@nectech.co.uk>
To:        "'FreeBSD questions'" <questions@FreeBSD.ORG>
Cc:        "'goshik@binep.ac.ru'" <goshik@binep.ac.ru>
Subject:   RE: Leased line - SLIP or PPP?
Message-ID:  <084DD226F592D211988800A024AC583B02B76D@exchange.nectech.co.uk>

next in thread | raw e-mail | index | archive | help
Hello,

This could be probably a problem with the UART in PC B. Maybe it's output
drivers are not designed to drive a 1km long cable, and the output rise and
fall time is suffering as a result. Slowing the data rate down would help in
this respect. Maybe the old CISCO's serial port had proper line drivers on
it's outputs.

Also, maybe the errors are due to the RX buffer's overflowing. With software
flow control (XON/XOFF), you have to be very quick to send the XOFF
character to the other modem. At this high data rate, maybe the RX buffer
has overflowed before the XOFF has been sent? You could fix this by setting
the RX IRQ trigger level to a lower value (if possible). 

I doubt if adding error correction or checking would get you any more data
down the pipe, although you might improve the reliability. You can't get
something for nothing, no matter how clever the software is.

It's hard to say what the problem is at the moment because you've changed
two things at once (ie. the UART and the data rate). Change one thing at a
time and it should become clear.

Your English is very good by the way, a lot better that some 'native'
English speakers!

Hope this helps,

Jeff


>Good time of day all.
>
>I have some problems with the connection of our LAN
>(3 Ethernet segments) to ISP's backbone.
>The connection is:
>
> 486DX2/66  +---+    leased line     +---+  486DX4/100
> 16MB       | A |--------------------| B |  8MB
> 3*3C509B   +---+     ca. 1 km       +---+  3C509B
> FreeBSD    ST16550                 16550A  FreeBSD
> SLIP                 MTU=296               SLIP
>
>Modems used are 115.2 kbps asynchronous ones without
>hardware flow control/buffering.
>
>When point B was Cisco's 2514 AUX port, speed was 38.4 kbps 
>only, but SLIP worked fine (netstat reported from 1e-4 to 1e-3
>of errors on input).
>
>But then we decided to change B to PC router to increase
>speed to 115.2.
>
>Now I'm getting from 10% to 20% of errors on input packets,
>but the worst is that connection acts with large delays.
>
>Looks like the line is rather noisy and the system spend much
>time retransmitting bad packets. As I recall, such a behaviour
>was mentioned in RFC1055 as a "feature" of SLIP (am I wrong?).
>
>Questions:
>1. How can I introduce some error correction / flow control
>   in this situation?
>2. If 1 is not possible, will the replacement of SLIP
>   with PPP be helpful?
>
>Note that it is not possible to change hardware/line
>in the near future, so I must play only with software.
>
>I suspect this is RTFM, but I'm quite new with UNIX
>and possibly you can give me some tips or pointers
>about SLIP or PPP usage over leased lines?
>
>Thank you for your time,
>Igor
>
>Sorry for my poor English


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?084DD226F592D211988800A024AC583B02B76D>