On Tue, 21 Mar 2000 18:07:14 +0100, Michael Polak wrote:
> So finaly I started with fine tuning of Arachne perforamce under the
> same old DOS, instead of dreaming about Linux port (but not forever!!)
> I have noticed, that with the same CORE.EXE, Arachne installation on one
> partition of my disk (C:) needs 10 sec to load one page, while
> installation of different partition (E:) needs 13 sec to load. I tried
> all possible methods to accelerate installation on E: drive:
> 1) I ran DISKOPT - it helped a little (cca 1 sec per page), but not
> enough...
> 2) I decided to modify Arachne to write .htt files to CACHE\HEADERS
> instead of directly CACHE\. This helped a little.... another second
> (the 10 vs. 13 ratio is AFTER all my fine tuning...). In this phase,
> number of files in CACHE on C: drive was higher, than number of files
> in CACHE on my E: drive
> 3) I reduced number of files in Arachne working directory to
> be smaller than 128, because I have heard this can help FAT filesystem
> used by DOS to run faster... it had absolutely no effect
> 4) Becoming totaly hopeless, I have started again, with totatly clean
> cache. Page load times from cache are now: 7 sec on drive C:, 11 sec
> on drive E:
> Both C: and E: use NWCACHE write behind disk caching. The differences
> are that C: drive uses 4 KB clusters, while E: only 2 KB clusters (to
> save disk space). Is it possible that my C: drive is faster because it
> uses sectors with smaller absolute distances than E: drive (?)
> I am going to compare load times with RAMdisk only installation, but
> I am surprised how big the speed differences can be even on the same
> PC, same operating system and same hard drive.... maybe this may be the
> reason why some users are enthusiastic about speed, while others
> complain ?
> BTW, the page I was testing is http://ctk.ceskenoviny.cz
> --
> Michael Polak: [EMAIL PROTECTED]
> Arachne Labs: http://arachne.cz/
> My mobile phone - up to 160 characters: [EMAIL PROTECTED]
Michael:
Surely enough, I am no DOS expert but a long time ago I didn't like at
all how my beloved Opus Discovery (ZX Spectrum drive) loaded the full 48k
RAM in about 25 seconds, while the rival Disciple used 3-4 secs. to do
the same.
So I tweaked the sector size and could bring the time down an impressive
40-50% going from the 256-byte standard Discovery sector to a 1024-byte
one.
I think the cluster size has to make a difference, I just am not sure
how much of it.
Hope this helps.
Enrique Baraibar
-- This mail was written by user of Arachne, the Ultimate Internet Client
WWWWw World Wide Web Without william
-- Arachne V1.50;s.r.c., NON-COMMERCIAL copy, http://home.arachne.cz/
