Subject: Re: misc questiosn about sun3
To: schaecsn <schaecsn@gmx.de>
From: Logan Bowers <logan@datacurrent.com>
List: port-sun3
Date: 07/27/2001 19:41:40
I'm assuming that since you only have 4MB of RAM you're using an NFS swap
file.  If that is the case, You cuold be seeing such a huge amout of
wasted time due to excessive thrashing that is magnified by the fact that
your swap file isn't local.  If you can, try moving the RAM from
your 3/110 into your 3/50 and see if that reduces your system time.  Also,
you might want to see how much time top takes when nothing else is
running.  The amount is non-trivial.  =20

Oh yeah, and I plan to use my 3/60s as dumb Xterms for my Linux machine
(only a 300Mhz AMD).  They're pretty effect for that; except that mozilla
can crash the Xserver.  I think the sheer complexity of the mozilla
window causes the Xserver to choke until eventually its buffers overrun
(from not processing data fast enough).  Of course this is just
speculation, I'm not definitely sure why.  When I get back to school in a
month, I hope to post an image of the root dir I use to boot the machine.
It's running NetBSD and X11R6. =20


Logan Bowers

"First they ignore you. Then they laugh at you. Then they fight you. Then y=
ou win."
--Mahatma Gandhi

On Fri, 27 Jul 2001, schaecsn wrote:

> Hi,
>=20
> I have gotten 2 sun3 computers: a 3/50 and a 3/110 with 4 MB or RAM
> each.
>=20
> * So far I have set up the 3/50 to boot diskless. It's not a speed
> demon. really, no at all! Is it possible that older kernels (older
> versions of netbsd) are faster on these old machines? or doesn't it
> matter?
>=20
> * Is somebody in the Bay Area interested in trading the 3/110 with
> another obsolete (perferably netbsd capable) system?
>=20
> * when compiling a program the CPU usage is around 10% User, 60% System
> and 30% Idle. My interpretation is that most of the 60% are used moving
> data over the ethernet via NFS from the diskless cient to the server and
> vice vera. since it takes some time to move it around the CPU has to
> wait 30% of the time and can compute the actual problem for 10% of the
> time. Is this a reasonable interpretation? If so, diskless throughput is
> just 10%!
>=20
> * right now I compile the kernel by myself. How long does it take, 1
> week?
>=20
> * that guy gets very hot after a couple of hours. what do you think,
> should I let this guy catch his breath and cycle him down for a while
> each now and then?
>=20
> * Is somebody actually using a 68020 based sun for something except
> showing his friends?
>=20
> - Stefan
>=20
> --=20
>    ___  ___       __         |   Name: Stefan Sch=E4ckeler
>   /  / /  /____  (__)__  ___ | E-mail: schaecsn@gmx.de
>  /  /_/  /  _  \/  /\  \/  / |    day: Solaris 2.5.1, HP-UX 10.20
>  \___,__/__//__/__/ /__/\__\ |  night: slackware 3.0, netbsd-alpha 1.5
>=20