*BSD News Article 4900


Return to BSD News archive

Newsgroups: comp.unix.bsd
Path: sserve!manuel!munnari.oz.au!hp9000.csc.cuhk.hk!uakari.primate.wisc.edu!caen!uunet!psgrain!percy!percival.rain.com!nerd
From: nerd@percival.rain.com (Michael Galassi)
Subject: Re: bsd386 slip xfer rate disappointing.
Message-ID: <BuFMzv.383@percy.rain.com>
Sender: news@percy.rain.com (News maintainer)
Nntp-Posting-Host: percival.rain.com
Organization: Percy's mach, Portland, OR
References: <3542@ra.nrl.navy.mil>
Date: Fri, 11 Sep 1992 21:14:18 GMT
Lines: 30

solsman@ra.nrl.navy.mil (Mark Solsman) writes:


>Is everyone getting the same response?  slip from 386/25 running bsd to
>a 486/50 running slip8250 & tn3270 yeilds about .8k/second???

>funny, I am using slattach /dev/com2 38400, that should be damn faster than
>800 bytes/second...

You bet...

>also, as far as the silo overflows , I seem to only get them when I 
>have a packet collision.  dont type anything until after xfers, and turn
>off the damn hashing, and you dont get any overflows...  this maybe a side-
>effect.

Colisions indicated by netstat -i are not collisions, by definition you
can't have collisions on slip links.  The collision counter is incremented
(counter intuitevly) when the output routine (slstart I think) drops a
packet instead of putting it into the com port's ring buffer.  They say
this is to improve interactive responsivenes and that tcp will take care
of it but...  I've got a hacked if_sl.c which I'll mail to those who know
enough to look over the changes and decide if they are indeed good.  End
result of these changes here is that my machine routes between two 38.4K
slip, one 19.2K slip, and one ether with very few errors (this is a 386/25
w/ 2meg ram & 1 floppy).

-michael
-- 
Michael Galassi -- nerd@percival.rain.com