f



SCO OpenServe 6.0.0 Filesystem MUCH Faster Than 5.0.6

The speed of the SCO OpenServer 6.0.0 filesystem is quite
noticeably faster than I enjoyed on OpenServer 5.0.6.

How do I know?  I donno, but it feels and seems so.

I didn't expect this, so I didn't make these tests while
still on 5.0.6, but I KNOW I have a lot more files and
directories on 6.0.0 than I did on 5.0.6.

The Verify of my Last Master Backup shows what is on my systen:

     255191 files were encountered
        167 files were excluded
          9 files were not checked because the hard disk file
            had been modified by a program since it was backed up
          2 files were not checked for various reasons
      28870 special files (links, directories, devices, etc.)
     226143 files verified successfully

Master Backups using LONE-TAR(tm) to DVD+RW media ran at about
70-75 MB/min on OSR 5.0.6 but run at 175+ MB/min on 6.0.0, and
if I turn off compression run at 260+ MB/min.

To see how long it takes to search for a non-existant file:

[trebor] ~: time find . -name XPMNYTRE

real    0m58.59s
user    0m1.72s
sys     0m10.96s

And to see how many directories had to be read:

[trebor] ~: find . -type d | wc -l
   23745

Bob

-- 
Bob Stockler  +-+  bob@trebor.iglou.com  +-+  http://members.iglou.com/trebor
Author: MENU EDIT II - The BEST Creator/Editor/Manager for filePro User Menus.
Fully functional (time-limited) demos available by email request (specify OS).
0
bob2838 (101)
11/25/2005 4:01:36 PM
comp.unix.sco.misc 3925 articles. 0 followers. Post Follow

3 Replies
1060 Views

Similar Articles

[PageSpeed] 13

On Fri, Nov 25, 2005 at 11:01:36AM -0500, Bob Stockler wrote:
> The speed of the SCO OpenServer 6.0.0 filesystem is quite
> noticeably faster than I enjoyed on OpenServer 5.0.6.
> 
> How do I know?  I donno, but it feels and seems so.
> 
> I didn't expect this, so I didn't make these tests while
> still on 5.0.6, but I KNOW I have a lot more files and
> directories on 6.0.0 than I did on 5.0.6.
> 
> The Verify of my Last Master Backup shows what is on my systen:
> 
>      255191 files were encountered
>         167 files were excluded
>           9 files were not checked because the hard disk file
>             had been modified by a program since it was backed up
>           2 files were not checked for various reasons
>       28870 special files (links, directories, devices, etc.)
>      226143 files verified successfully
> 
> Master Backups using LONE-TAR(tm) to DVD+RW media ran at about
> 70-75 MB/min on OSR 5.0.6 but run at 175+ MB/min on 6.0.0, and
> if I turn off compression run at 260+ MB/min.
> 
> To see how long it takes to search for a non-existant file:
> 
> [trebor] ~: time find . -name XPMNYTRE
> 
> real    0m58.59s
> user    0m1.72s
> sys     0m10.96s
> 
> And to see how many directories had to be read:
> 
> [trebor] ~: find . -type d | wc -l
>    23745
> 
> Bob
> 
> -- 
> Bob Stockler  +-+  bob@trebor.iglou.com  +-+  http://members.iglou.com/trebor
> Author: MENU EDIT II - The BEST Creator/Editor/Manager for filePro User Menus.
> Fully functional (time-limited) demos available by email request (specify OS).

The 6.0.0 filesystem IS significantly faster than 5.0.x, but that is
probably not the reason for your faster backups in this instance.

The ATAPI driver in 5.0.x runs only in PIO mode (and not a very fast PIO
mode at that), while the 6.0.0 driver runs in DMA mode, which gets you a
lot closer to the theoretical performance limit of your DVD+RW device/media
combination.

-- 
Tom
---
 D. Thomas Podnar
 tom@microlite.com         http://www.microlite.com
 Microlite Corporation     724-375-6711 Voice
 2315 Mill Street          724-375-6908 Fax
 Aliquippa PA 15001-2228   888-257-3343 Toll Free Sales
-------------------------------------------------------
Developers of Microlite BackupEDGE
0
tom4811 (60)
11/26/2005 7:03:56 PM
Tom Podnar wrote:

> The 6.0.0 filesystem IS significantly faster than 5.0.x, but that is
> probably not the reason for your faster backups in this instance.
> 
> The ATAPI driver in 5.0.x runs only in PIO mode (and not a very fast PIO
> mode at that), while the 6.0.0 driver runs in DMA mode, which gets you a
> lot closer to the theoretical performance limit of your DVD+RW device/media
> combination.

Hmmm, the OSR5 IDE driver intends to run in the highest PIO mode
supported by the hardware.  A few years ago I remember observing that
then-current ATAPI CD/DVD writers only supported up to about the
low-middle of the PIO speed range.  If you're seeing OSR5 still using a
slow PIO mode with ATAPI CD/DVD writers, it is almost certainly because
the hardware still limits itself to that speed (for incomprehensible
marketing reasons, or just because they never saw a need to go faster
since "everything" uses UDMA...)

>Bela<
0
filbo (325)
11/26/2005 9:33:45 PM
On Sat, Nov 26, 2005 at 04:33:45PM -0500, Bela Lubkin wrote:
> Tom Podnar wrote:
> 
> > The 6.0.0 filesystem IS significantly faster than 5.0.x, but that is
> > probably not the reason for your faster backups in this instance.
> > 
> > The ATAPI driver in 5.0.x runs only in PIO mode (and not a very fast PIO
> > mode at that), while the 6.0.0 driver runs in DMA mode, which gets you a
> > lot closer to the theoretical performance limit of your DVD+RW device/media
> > combination.
> 
> Hmmm, the OSR5 IDE driver intends to run in the highest PIO mode
> supported by the hardware.  A few years ago I remember observing that
> then-current ATAPI CD/DVD writers only supported up to about the
> low-middle of the PIO speed range.  If you're seeing OSR5 still using a
> slow PIO mode with ATAPI CD/DVD writers, it is almost certainly because
> the hardware still limits itself to that speed (for incomprehensible
> marketing reasons, or just because they never saw a need to go faster
> since "everything" uses UDMA...)
> 
> >Bela<

Doesn't matter who gets the blame.

2 years ago (2 Nov 2003) we did extensive testing on different machines,
and our results scaled with CPU speed. We used devices that had 
PIO Mode 4.

On our P4 2.67GHz machine back then, we topped out at 2830290 bytes/sec
or 161.95 MB/min with OpenServer, while our 4X Plextor PX-708A (the
fastest device we had at the time) under UnixWare 7.1.3 ran at
5456022 bytes/sec, or 312.2 Mb/min on the same machine.

Tape drives running on the ATAPI bus are also slow, and I suspect for
similar reasons.

All this information was provided to SCO back then, and you saw it
as part of your duties at that time.

I can run a REV vs REV test, since it is currently the fastest ATAPI
device with an MMC bios, but it will just confirm the findings.

The PIO implementation of 5.0.7 is just slow. So is the USB implementation,
although not as bad. But given the mandate to work on this problem or
get OSR6 out the door, I'd have done exactly that same thing as SCO and
gotten OSR6 out the door.

Tom Podnar
0
tom4811 (60)
11/28/2005 10:39:29 PM
Reply: