[EMAIL PROTECTED] wrote:

Jacob Anawalt wrote:

>   [EMAIL PROTECTED] wrote:
>
>>Hey guys (are there any girls on this list - hmm)
>>Which filesystem would you recommend for a biggish (10 gig) partition ?
>>At
>>the moment I'm using Ext2fs and I really have to use something else,
>>it's
>>just way to slow. So should i use Reiser or Ext3fs ? Or is there maybe
>>another fs I can use ?
>>In what areas of use does the Ext2 FS on your 10GB drive seem "way to
>>slow"?
>>
>When a file transfer starts my cpu jumps to 100% and my pc almost
>freezes up
>for a while before i can work again. It's horribly slow on read and
>write requests.
>While playing a movie and working on a xterm the cursor stops responding
>every now
>and then.

That's pretty bad. I suspect the disk and cpu are not talking optimally.
See below.

>>
>>Transferring a file from one folder to another?
>
>yes
>
>>With only one disk you're copying from disk, to MB and writing back to
>>the same disk - not the speediest of disk operations normally. It would
>>help to know if this partition contains all of Debian or if it is just
>>one mount like /home and if the rest of Debian is on the same disk or
>>another disk
>
>All of debian on 1 partition - I know this is not recommended, but I'm
>not
>in the mood to create a bunch of partitions and try and manage that.
>

That's ok. I have it all on one disk on this machine. It works just fine. I
didn't know what level of performance you were after.

>>Reading files off the disk?
>
>yes
>
>>Linux nicely pre-fetches data beyond your read request anticipating that
>>
>>you are going to need that data, thus speeding disk reads by reducing
>>the need to move the head around. Severely fragmented files may not have
>>
>>their data within that read-ahead section and miss out on that potential
>>
>>performance gain.
>
>I though ext2fs didn't fragment ? Seems like some of the linux-boys here
>at work don't know what they're talking about ;-).
>

Its not that it _doesn't_ever_ fragment. It handles fragmentation very well,
except for files that are constantly appended to - like large directories and
log files (or mailbox files for an IMAP account where the user never removes
the mail.) Some day when your mount count catches up and you're watching fsck
go on all 80gb you'll see the frag % report.


>>Files that are always appended to (like some tar
>>files, directories that keep getting new entries, or system logs) are
>>very likely to fragment. Files that are re-written get a clean slate and
>>
>>are not fragmented. Once the kernel reads something it keeps it in
>>memory if possible since that is much faster. If you don't have much
>>free memory, you are missing out on that gain as well.
>
>I've got 800 MB's of ram so i think it's not a memory problem
>

Um, ya. Unless you have some huge pic's open in Gimp and stuff.

>>I don't know if
>>the Resier FS handles fragmentation of appended files better, but any
>>file system will appear to work faster on cached data if you have oodles
>>
>>of memory.
>>
>>Some other situation?
>>Maybe the disk/controller isn't using DMA? What type of disk is it?
>
>It's a seagate Baracuda 80 gig, so it does use DMA. Are there any
>specific drivers
>I should load for DMA to work properly ?

The Baracuda 80G that I found is ATA/100. If the controller is talking to
it at ATA 100 and if the controller is using DMA then I'm scraching my head.

More likely it's not. Checkout the manpage of hdparam, and search around for
hd tuning. If hdparam cant set dma, it could be your chipset isn't being
recognized. Maybe some module needs to be loaded or built into the kernel.

>
>>I haven't used much of anything outside of ext*, so I can't give an
>>experiance based comment on Reiser. It's suppose to be pretty fast and
>>good at handling lots of small files and directories.
>
>Can you define small file for me ? Small as in < 5k / < 10 k
>

Sorry, I haven't looked that much into ReiserFS. Someone else mentions XFS and
while that may run faster, I think the real issue is your IDE controller.


>>The place where ext3 orders of magnitude faster than ext2 is when you
>>boot up after an improper shutdown/unmount. ext3 (which is basically
>>ext2 with journaling) can safely skip fsck for the most part. Ext2 you
>>get to sit there watching the file systems scan away. The bigger your
>>hard disk, the longer you'll sit and the faster ext3 would seem. :) In
>>other cases, because ext3 is writing it's journal to a disk every few
>>seconds, it could be a little slower. Since most systems aren't normally
>>
>>under intense I/O, this is usually unnoticable. You do have the option
>>of storing the journal to a different disk.
>
>So what you're saying is that I'm actually using the faster filesystem ?
>Barring disk checking now - which doesn't really matter anyway, it's
>not like that happens every day)
>
>From what I've read I'm beginning to think I've got some driver problem.
>

I think so too.

>>Ok, I've hit the end of any helpfull thought on the subject, now for
>>some related ponderings:
>
>You've been very helpfull thanx dude.
>
>
>Sorry for the bold italics, but I was getting sick of indenting with >
>line by line. Lotus Notes is probably the most stupid email clients
>around when it comes to that
>
>

Ya, that formatting was pretty horrid. You could probably do a better job
by editing the message in vim/emacs first. Infact I used VIM to try and 'fix'
the formatting. Hopefully it worked right. ;)





--
To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]





Reply via email to