TIP: Click on subject to list as thread! ANSI
echo: locuser
to: All
from: Paul Edwards
date: 1996-05-25 00:57:10
subject: hpfs

Well, what's the performance like after my upgrade?  Well,
first of all, when I write messages, and reply to them (which
brings up the external editor), it just snaps up really nicely,
instead of watching it load the editor.  I may have confused
the issue by running HPFS386 on and off though.  Rescanning
messages is virtually instantaneous, as the HPFS386 4 meg
keeps all the indexes required in memory.  That leaves my 
overnight tasks, when I do backups, compress messagebases, etc
etc, which used to take 1 hour, now it takes:

Daily maintenance took 1 hour before new MB
22 mins with new MB (still HPFS)
40 mins with new MB + HPFS386 w/- 4 meg cache
40 mins with new MB + HPFS386 w/- 2 meg cache

Which is a great shame!

I remembered that I got those sorts of results before, and 
didn't complete the investigation, but did remember the ballpark
I had got to.  That was sqpackp.

I have been able to reproduce the problem, and it goes like this:

With my current (5 meg) AVTECH, I get the following results on
running "sqpackp avtech":

hpfs 2 meg - 13 seconds
hpfs 1 meg - 13 seconds

hpfs386 2 meg - 32 seconds
hpfs386 4 meg - 32 seconds
hpfs386 8 meg - 32 seconds

The following sequence of commands:
copy /b avtech.sqd nul:
sqpackp avtech
STILL made it take 32 seconds.

The following sequence of commands:
copy /b trading.sqd nul: [to clear avtech.sqd out of the cache]
copy /b avtech.sqd nul: [about 3 seconds]
sqpackp avtech [about 4.3 seconds]

Makes my scanning time go from >twice as slow to about twice as 
fast!

As far as I can tell, the operation of sqpackp is as follows:
open avtech.sqd for read
open avtech.~~~ for write
selectively copy from avtech.sqd to avtech.~~~ [in my case, the
copy will be identical, as there is nothing to be packed in most
of my messagebases].
delete avtech.sqd
rename avtech.~~~ avtech.sqd

I do not yet have a theory to explain the observed behaviour.
I have played with bufferidle and maxage until I'm blue in the
face, but I can't find a way to make hpfs386 get the 13 seconds
I am looking for.  That 7.3 seconds clocked above, will only
work if I have the whole thing in memory [I used an 8 meg cache
to achieve that, on a 5 meg file].  At an estimate, half of my
messagebase data is in files > 8 meg.  

Anyone got any suggestions?  BFN.  Paul.
@EOT:

---
* Origin: X (3:711/934.9)

SOURCE: echomail via fidonet.ozzmosis.com

Email questions or comments to sysop@ipingthereforeiam.com
All parts of this website painstakingly hand-crafted in the U.S.A.!
IPTIA BBS/MUD/Terminal/Game Server List, © 2025 IPTIA Consulting™.