Re: [tlug] Disk I/O bottleneck: how to solve?

["Jonathan Q" <jq@example.com>]

On 9/4/2002, "Jean-Christian Imbeault" 

>The hard drive is already pretty fast (ATA-100 7,200 rpm).

Just one disk?  I take it there is no slave on the same cable?

>drives (I heard that ATA-133 is really only about 5% faster that ATA-100).

In the real world, I doubt it's actually possibly to achieve the transfer rate
of ATA-100, or even get near it most of the time, so I believe the claim about
133.  The speed difference you see would probably come mostly or completely
from faster seek and read/write times on the drive rather than electronic
differences.

>*could* go SCSI but since it is a much more expensive option it would need
>to be a big imporvement in speed.

A very fast seeking, fast writing Ultra-160 15,000 RPM drive would certainly
beat an ATA-100 X 7200 RPM hands down.  However, if the problem is coming from
too many reads/writes at the same time, you would see an improvement but not a
quantum leap.

RAID 0 or RAID 5 with several fast disks would help, but you're constrained by
lack of drive bays.  One possible option is to put in a U160 SCSI board and
have an external SCSI box filled with several very fast disks.  This does, of
course, require a substantial hardware investment.  If you have a fat budget, 
you could go full hardware RAID :-)

Jonathan


>
>I was wondering if the were any software options? My machine has a lot of
>free RAM and I was wondering if there was any way to make the kernel use
>more RAM as a disk buffer? My machine always has at last 200M of free RAM
>(total of 512M) so using that as a disk buffer could really speed things
up.
>
>Is it possible?
>
>If there are other solutions I have overlooked please let me know ofoucrse
>;)
>
>Jc
>
>
>
>_________________________________________________________________
>Chat with friends online, try MSN Messenger: http://messenger.msn.com
>
>
>**********************************************************
>TLUG server is hosted by Open Source Development Lab Japan
>http://www.osdl.jp/
>**********************************************************
>
>==========================================================
>To unsubscribe from this mailing list,
>please see instructions at
>http://www.tlug.jp/list.html
>==========================================================
>
>
>

GPG key: DF12B4EF (5399 C834 3ABB C3AF 610C  5345 D5D6 E6EA DF12 B4EF)
gpg --keyserver pgp.mit.edu --recv-keys  DF12B4EF