On Wed, Oct 03, 2007 at 04:24:59PM +0200, Damir Perisa wrote:
> the file itself is not fragmented. but the files all together are not 
> at the same place after a long time of usage. in the worst case the 
> harddrive has to do seek for every file instead of by one run read 
> the whole pacman database.
> 
> since this fragmentation happens only over a long time (years), most 
> of the users will not feel the optimisation immediately. it makes 
> also an implementation of a more sophisticated (complex) database 
> futile. (a mysql db would be much faster than accessing single files 
> i guess)
> 

I don't know if this is just an impression, but I find my hard disk awfully
slow for a few months.
The two main apps concerned are pacman and mutt, since they both deal
with a lot of files.

# sync; echo 3 > /proc/sys/vm/drop_caches

# time pacman -Qu
Checking for package upgrades... 
no upgrades found.
pacman -Qu  0,61s user 0,52s system 1% cpu 1:20,00 total

Well, actually, it seems like it's even slower using this drop_caches trick
than after a fresh boot, but I'm not even sure..
pacman-optimize didn't help at all.

This is with a laptop, core duo t2300, 1gb ram, 80gb sata disk, ext3
filesystem.
The disk is apparently a FUJITSU MHV2080BH.

Filesystem            Size  Used Avail Use% Mounted on
/dev/sda5             9.2G  6.2G  2.6G  72% /


I also have an older desktop, p4 3.2ghz, 1gb ram. But it has two "WD 36.7Go
10000 RPM S-ATA (Raptor)" in raid 0. ext3 filesystem too.
So it's supposed to be faster, but the above command run in 3 seconds.
So 27x faster? Isn't that a lot?

Also, I don't remember this laptop was that slow when I got it one year ago.

_______________________________________________
arch mailing list
[email protected]
http://archlinux.org/mailman/listinfo/arch

Reply via email to