Considering how much slower hard drives are than memory, any instance where memory can take the work load of the hard drive will give an improvement of several orders of magnitude. I would say don't worry too much about the small speed difference between 7200 and 10-15000 rpm drives. Instead, invest in more memory and concentrate on system set-up.
If two large blocks of different data are being read/written concurrently, the drive head will keep scanning accross the disk surface between the two locations (thrashing). This is extremely inefficient whatever mechanical system you use. When loading an application on a memory-constrained system, thrashing will occur as the drive head scans from the application binary to the swap space and back.
The Linux kernel caches recently used disk data in memory not used for applications. Take a look at the memory used by applications. Take steps to tame over-greedy apps to leave more cachable.
If using swap space, use a separate old, redundant drive as swap. (you should have enough memory not to use swap, but if you do, this is an easy way of taking load off the main drive.). This will stop thrashing at application load-up.
If you have some power users which access large files concurrently (eg video editing), try to distribute these users home directories across different physical drives.
Set more read-ahead caching where the drives are thrashing and other methods can't be used and check write cache is enabled. Also consider setting noatime on file systems. atime writes the access time to the file system each time a file is accessed. Disabling this reduces load.
jef peeraer wrote:
On Wednesday 12 May 2004 06:08, Sudev Barar wrote:
On Wed, 2004-05-12 at 07:29, Chris Berry wrote:
How much do you think disk speed affects the latency, and is it worth it to use 15k rpm scsi disks instead of 10k rpm. Will we notice serious speed loss using IDE for 2 or 3 clients, or only when we start loading up more?
I am running one system on IDE one on SATA and more on SCSI. Always with more clients coming in (beyond 10+ concurrent) latency starts becoming apparent and irritating as more clients become concurrent. Two systems that are on SCSI today I used RAID0 on two IDE's to solve part of the problem and although there was improvement in read / writes as shown by hdparm it was still not good enough. SATA is much better but no where near SCSI. Lot of discussions on this can be dug up in archives of the list. As far as two three clients as proof of concept the current hardware is okay.
what about these 10000 rpm sata drives . will they be sufficient for 10 + loads ?
-------------------------------------------------------
This SF.Net email is sponsored by Sleepycat Software
Learn developer strategies Cisco, Motorola, Ericsson & Lucent use to deliver higher performing products faster, at low TCO.
http://www.sleepycat.com/telcomwpreg.php?From=osdnemail3
_____________________________________________________________________
Ltsp-discuss mailing list. To un-subscribe, or change prefs, goto:
https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help, try #ltsp channel on irc.freenode.net
