X-PMC-CI-e-mail-id: 9563
Leonard wrote:
>It is very difficult to find exact weight of dogs in different species :-)
>
>I am dealing with the real world.
Exactly. So we are all.
What you obtained is simply the "perceived" speed of copying 9 MB of
files on respective systems:
The time for such transfer is a function of various factors (real
life).
What you showed was
time(SCO OpenSErver, DPT, SCSI disk, file system used,
CPU speed, RAM speed, RAM amount, ...) for running
a particular script that copied a certian disk directories
on that SCO openserver PC
was longer than
time(linux, IDE controller, IDE disk, file system used,
CPU speed, RAM speed, RAM amount, ...) for running
a particular script that copied a certian disk directories
on that linux PC.
I don't see from this simple data that IDE disk is better than SCSI
because there are too many factors in the time() function above that
are not properly controlled in your experiment.
You need to control the various factors that can
affect the value of time() function above to the extent
that only disk controller and disk itself matters in the measurement.
(Then we can compare "apple" and "orange", maybe.)
Unless such data is available, it only shows that YOUR linux PC took
shorter time than YOUR SCO PC to finish the script for copying
particular directories on both systems in the way you ran them. There
is no question about it. I am sure everyone on the list agrees about
it based on your reportd data.
But as the various posts showed that it had nothing to do with the
disk controller and raw disk performance issues.
Finishing of the script doesn't even mean that the data was copied to
the disk via the disk controller at all. (It was probably written to
the main memory file buffer inside linux OS and not copied to the disk
until sync, eh, update runs again.) So this data you obtained has
nothing to do with IDE vs SCSI argument. That is what the various
posts tried to convince you.
When people speak of performance issues in technical circles, usually
people take care at least to decrease the effect of these "other"
real-life paramaters.
Even limiting the scope to computers, if you are in doubt, consult
SIGMETRICS publications of ACM at a local college library or on ACM
web sites. Or check out some titles of the proceedings at
http://www.usenix.org (or find them at local college library). Even if
you can't access the main text of the articles, etc. try looking up
the web pages of the authors mentioned in the titles in the web pages,
and proceeding papers, etc.. This will give you some ideas about the
controlled performance measurement criteria.
I think this criteria also holds true in other engineering
discpline(sp?) such as combustion engine performance, etc..
Unless you can come up with better data in a more controlled
environment, let us kill subject once for all.
Sorry, I couldn't resist to chime in.
OH, by the way, I am one of those dinosaurs using
50-pin SCSI disks on two of my PCs happily :-)
currently DC390, Symbios controller from a company called IOI,
BusLogic 9xx (forgot the name),
and used VLbus BusLogic contoller and ISA AHA1542 before.
If only more reliable data is in regarding IDE and SCSI issues.
I haven't bothered to look at the large IDE disks considering the
elecrical characteristics of the IDE cable (very unreliable at high
speed.) But lately, I am seriously looking at the large IDE disks due
to budgetary reasons (a good cause for a home PC), and I am afraid the
only useful post I saw in this thread was from Alan Cox syaing that
for a single user PC, maybe SCSI is not worthwhile considering the
price difference. I kind of agree and tend to place the source file
archives that I downloaded from net on IDE. But all my often used
files and referenced database files are on SCSI disks (i think).
But more carefully measured data comparing the latest IDE and latest SCSI (or
old-fashioned 50-pin SCSI) would be welcome.
I don't need a speed daemon for my home PC, but don't want a very
slow disk under real-world condition such as kernel re-compilation
either.
--
Ishikawa, Chiaki [EMAIL PROTECTED] or
(family name, given name) [EMAIL PROTECTED]
Personal Media Corp. ** Remove .NoSpam at the end before use **
Shinagawa, Tokyo, Japan 142-0051
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
