[Sig] SuSE Linux LiveEval First Observations

[Randy Kramer]

Just thought I'd share this with anybody who's interested.  Even though
it sounds like a disappointing experience, I still think the SuSE 7.0
LiveEval with KDE 2.0 will be a good introduction to Linux for Windows
users -- we might have to specify a minimum machine for reasonable
performance, and the machine described below is below the minimum.  I'm
not sure where the biggest problem is, but I suspect the problems rank
in this order (from most significant to least significant): speed of the
CD-Rom drive (or use of the CD-Rom instead of installing to the hard
drive), amount of RAM, speed of the processor.

I tried the SuSE 7.0 LiveEval from the CD on a 133 Mhz machine with 47
MB of RAM and a 4X CD drive.  It put a 100 MB "working" file system
(mounted as /home) and a 100 MB swap file on drive C:.  (It tries to put
these on drive C:, if the C partition is not big enough, I'm not sure
what happens).  There is apparently a way to force more of the system on
to drive c to make what they call a "permanent" installation (without
partitioning the disk, AFAIU).  I have not investigated that.

The system is slow, some slowness is probably attributable to the lack
of RAM (they recommend 128 MB), some to the slow processor, and some to
the slow CD-Rom.  I will try it later on a machine with a 36X CD-Rom, 96
MB, and a 200 mhz processor (which Mandrake "describes" as almost 400
bogoMips).  

(I don't know how reliable a measure of CPU (or system) speed a bogoMip
is, but I'm assuming it has some validity in comparing relative
performance of processors (or systems?) -- a 400 bogoMip processor would
presumably be about twice as fast as a 200 bogoMip processor.  I don't
know if a bogoMip only addresses processor speed or if it attempts to
consider the effect of available memory and access speed.)

Installation is agonizingly slow.  I typically answered a question, then
switched to another computer to do something useful, and then switched
back.  The first time I tested my video configuration I did not measure
the time the screeen was blank.  The installation warns (appropriately)
that if you test the video configuration and it doesn't work right,
quickly hit escape to avoid the possibility of damage to your computer. 
Well, that will be a toughie.  I went back and did the test a second
time and timed it.  The screen was blank for almost 30 seconds, and it
took almost 1:30 minutes till the test picture appeared, and, I think it
worked faster the second time than the first time.  (I'm guessing that
some of the necessary software was still in RAM and did not have to be
loaded again from the CD-Rom.)  If we give this to newbies, we should
try this install on a number of different machines, and give them an
idea of estimated video test times on different machines.

I don't have a modem on this machine, so could not test the dial up
setup.  It did find the Ethernet card and set it up, and gave me the
opportunity to set the IP, the host (domain) name, the default search
domain, and the DNS addresses.  It did not give me a chance to set up
the /etc/hosts file with the names of other machines on my LAN (which
caused problems later when trying to set up the printers -- see next
paragraph).

There is no printer connected to this machine.  I tried to set up a
printer on a Windows machine using Samba (one of the setup options,
labeled, I think, Windows / Samba).  It did not work, and I believe it
did not work because it could not resolve the host name to an IP address
(because I did not have the opportunity to enter that information into
the hosts file).  Also, at some point during a printer install for
Samba, it attempts to ping the printer host.  On one trial I put the IP
address in instead of the host name, and it seemed that the ping was
successful.  (The installation process is not very voluble.  It seemed
to hang during the ping whether it was successful or not.  When I
"cancelled" the ping I got a different result after pinging the IP
address then I did after pinging the host name, which made me believe
the ping of the IP address was successful.  (This is somewhat confusing
unless you do it -- you're in a graphical install and the ping is done
behind the scenes, you just see a text box with a message like "Do you
want to ping -c Office6", with choices yes and no -- you don't see the
command line or the actual results of the command.)

I could not configure my sound card, but I didn't expect to because the
sound card is built into the motherboard and no other Linux distribution
has been able to automatically install the sound card.  I'm not that
familiar with the possible options during installation of a sound card,
but there did not seem to be a chance to select the basic type of card,
like SoundBlaster, Gravis, or whatever.  It may be that it assumes that
the sound card is a soundblaster or clone, and only allows selection of
things like IRQ, memory, DMA, etc.  (Since Dan uses SuSE, it would be
interesting for him to try this install and see if it is the same
install used for the full distribution -- maybe it has been pared down
for some reason).

After the installation was complete, the system is still very slow, but
I can detect two separate components to the slowness.  When I call up a
new application, I attribute the slowness to the CD drive.  When I
scroll a window, I attribute the slowness and jerkiness to insufficient
RAM (I think).  (It is less jerky than KFM is under Mandrake 7.0 on this
same machine, but with KFM I could open 6 windows with 47 MB of RAM and
get decent response when switching between windows.  I don't believe I
will be able to get a decent response with 6 open windows with KDE 2.0
(I suspect it's more a result of KDE 2.0 vs. 1.2 than of SuSE vs.
Mandrake.)

To judge performance, I started 4 konqueror windows, 1 kmail window, and
1 terminal with two "subwindows", one running top and one idle.  Suse
also opens something they call the X-console (I haven't see this on any
other distribution so far, though it may be there) -- it has no prompt
-- I don't know what it's used for.  (And it never displayed anything
while I was opening or closing windows -- maybe it displays error
messages if an error occurs in X.)

With that many windows open, performance was sluggish, so I started
closing windows looking for a point when performance improved.  (To test
I tried two different approaches.  Several times I would cycle through
all the open windows one after the other.  Then I might try cycling back
and forth between just two windows.  I expected that I would get better
performance with that approach when both windows fit in RAM (i.e., they
didn't have to swap to the swap file).  With this number of windows open
, I could not make that distinction.  (The simple terminal windows did
open much faster if I alternated between them, but the konqueror windows
would not (if I cycled between a pair of them) -- maybe there is not
enough available RAM with a 47 MB system to hold even two konqueror
windows in RAM?)

I started closing windows one at a time to see when performance became
acceptable.  I really didn't notice any improvement even when I got down
to only one open konqueror window (plus the X-console, which I could not
close).  With only one open konqueror window I simply alternately
"iconicized" and maximized the window.

So, then I opened more windows, trying to see when I noticed a
significant performance decrease (below even the level of sluggishness I
was already seeing).  

I opened 6 konqueror windows, and 1 terminal (with two subwindows, one
running top).  Then two more konqueror, followed by two more.  By that
time I had 10 konqueror windows open, and it was somewhat more sluggish,
but I didn't find the significant increase in sluggishness I found under
Mandrake 7.0 with KDE 1.2.  I didn't go any further, I'm going to reboot
this machine to Mandrake 7.0 (which is still loaded) and refresh my
memory.  (PS: It's clear that KDE has improved the repainting in KDE 2.0
-- in 1.2, repainting was a circus, repainting some portions of a window
two, three, or more times, and uggh, it was just ugly.  Maybe improving
the repainting has reduced the increase in sluggishness (by reducing
memory usage?? -- I don't see the connection there.)

(All of the testing described above was with pages already loaded. 
Loading pages from the Internet was ourageously slow, but, believe it or
not, I suspect the system was accessing the CD-Rom while loading pages. 
Why, I don't know.  I am convinced of this because the green light on
the CD-Rom would commence flashing when I started to load a different
page, and remain flashing until the page was loaded.  Loading a terminal
or top or whatever also set the drive light to flashing.)

I welcome comments from anyone, but especially from anyone who tries to
run SuSE LiveEval on another system, with more or less RAM or a
different speed CD-Rom (or anyone who installs it "permanently" to the
hard drive.

Randy Kramer