Michael Polak wrote:
>
> Dne Fri, 14 Jul 2000 00:24:53 +0200, [EMAIL PROTECTED]
> (arachne-digest) napsal:
>
> > My previous time on the P90 to reload that interview from cache (offline) by
> > using the back key from another URL was 3 mins 13 sec. 193 Seconds total.
> > My time using DOS 5.0 to reload from cache as above was 26 Seconds !!!!!
>
> > That's a 700% improvement. And my HD doesn't sound like a machine gun now.
>
> I think disk caching has lot to do with this... and also disk
> fragmentation and so. There can't be 700% improvement between DOS 5 and
> DOS 6...
It seems that there CAN be that big improvement at least in the time it takes
to INITIALLY access many random files or to DELETE them.
Michael, the two above tests were done on the same machine with the SAME
disk caches. Nothing was changed except DOS 5 was booted from floppy with
a copy of the same config.sys as on HD except the comspec was pointed to A:
The autoexec simply called the autoexec on HD. And the same problems were
observed with ALL DOS 6 versions. 6.20,6.21,6.22
This following PAIR of tests -both DOS 6.2 and 5.0- were done with a smartdrv
cache of 2MB.
Quote:
My previous test with speedchk:
My P90 using DOS 6.2 30 sec create/copy, 41 total = 11 sec delete ***
My latest test with speedchk:
My P90 using DOS 5.0 30 sec create/copy, 31 total = 1 sec delete !!!!!
/Quote.
Today I have new info which helps explain the bad times seen on the slashdot
interview where there were about 8464 atoms.
My 340Mb HD has 2 partitions : "C" is about 103 Mb and "D" is about 230 Mb.
Therefore, the "D" partition is at least about 1/3 of the disk further away
from track 0 (the recalibrate point) than partition "C".
To recalibrate from partition "D" should take much longer on average than "C"
I reran the above "SPEEDCHK" on drive D:
Using DOS 6.2 on drive D: 35 sec create/copy, 51 total = 16 sec delete *
Using DOS 5.0 on drive D: 35 sec create/copy, 36 total = 1 sec delete !!!
Drive D: takes much longer than drive C: (see above) as expected.
The disk has a track to track step time of 1mSec with a full stroke time
of 20.6 mSec for 311 cylinders.
1/3 full stoke is approx 7 mSec, so to recalibate and return from 1/3 out
should take a MINIMUM of 14mSec. and to do that 228 times (there were 228
files) would take a MINIMUM of 3 extra seconds. The speedchk bat was NOT
at the start of drive D: but followed Win 3.11 + 2xNetscape + 2xLinux.
Drive D: took 5 more seconds than drive C: so that is quite reasonable.
Now, the same math for accessing 8464 atoms just ONCE each at a position of
about 40% of drive C: is 20.6 mSecx.33x.4x2 = 5.4 mSec in and out times that
8464 atoms equals only 46 seconds so I guess either this is totally bogus or
Arachne accessed each atom about 4 times.
Doesn't matter. DOS 5.0 did in 1 second what DOS 6.2 did in 16 seconds.
- Clarence Verge
--
- Help stamp out FATWARE. As a start visit: http://home.arachne.cz/
--
