Re: PPP slow between Linux and Solaris - why is it so?
Mon, 15 Jul 1996 12:53:42 +0100 (BST)


Dave Cole <> wrote:
> I remember someone mentioned that Solaris was retransmitting packets too
> quickly or something causing >50% of bandwidth to be used up by duplicate
> packets. I got the impression that Solaris was to blame. Is this
> correct? In any case, I do not remember anyone ever mentioning a fix /
> trick / work around to get the throughput back up to what it should be.

The problem I had was exactly the opposite, AFAIK. The link was only going
about 1/3 of the time. I didn't have a copy of tcpdump for PPP at the time
and I've since moved and no longer use dialup.

- From what I managed to work out, the Xylogics box was buffering a quite
considerable amount of data. It then seemed to panic and send quite a few
ICMP slow-down messages[1] back to the SunOS/Solaris host. (Although not being
root on the box in question, I can't be sure) Exponential backoff would then
mean that the delay shot up to about 30s, before gradually coming down again.
The whole thing would then repeat.

The problem is a Solaris/SunOS one as far as I can work out: it overreacts to
the ICMP messages. The solution I found was to write a bit of code to feed FTP
data to the remote end at a certain speed. (1k per second) which increased
throughput quite dramatically.

I don't know if the patches will fix the problem: Next time I'm on a Linux
box with a modem, I'll dial up and use one of the Solaris 2.5 machines here
to ftp data back into my Linux box and see what performance I get

All the above could be rubbish, but it's what I *think* was going on.

[1] I forget the proper term, but I think you know the ones I mean.

* * Ryan Richard O'Connell 0958 204802 *
* * Finger: *
* I'm looking on the bright side, I wear it like a bruse, -----=====*****
*****=====----- I've never loved Elvis and I've never sung the blues. *

Version: 2.6.i