solarflow99 wrote:
On Mon, Jun 9, 2008 at 11:28 PM, Sharpe, Sam J <[EMAIL PROTECTED]>
wrote:
At some point, solarflow99 wrote:
On Mon, Jun 9, 2008 at 8:45 AM, solarflow99 wrote:
The first one is a SATA 12 raid 6 array, the second one is a SATA 6 x
750GB
raid 5 array. I can understand 12 drives should provide more iops, but
this
makes no sense.
Sorry, too many variables. You are using at least:
-different OS (el4 vs el5)
-different drives (platters/firmware/cache?)
-different RAID levels
Also perhaps:
-different RAID cards (model/cache/bus/driver)
<https://www.redhat.com/mailman/listinfo/rhelv5-list>
really now, the hdparm tests show the throughput; this is a case of real
world results being completely opposite of hdparm tests. it still makes
no
sense.
Err. hdparm is an artificial benchmark - nowhere does it guarantee
equivalency to real-world metrics.
You haven't considered the disk-in-memory cache for one, are these systems
exactly equivalent? Basically, comparisons of this sort are only valid when
everything else is the same.
The same system connected to two different disk arrays is a valid
comparison. Two systems using two different arrays are not, because any
equation derived has too many variables to be solvable.
it seems hdparm is not as useful as I thought then, I wonder if you have
any preferred programs for disk IO benchmarks? It still strikes me as
surprising that the results could be so drastically different, the raid
controllers were not the same brand, but it was fairly similar hardware,
nothing I could account for that much difference.
bonnie++
hdparm does a straight-line test of some number of contiguous sectors.
If the box does its own caching (as I've seen in decades past on
mainframes), it can be "fooled."
Any benchmark can be realistic or not, depending on how it reflects
_your_ usage. A disk array with a Gbyte of cache will perform better if
the application stays mainly in the cache than if it really needs two.
Even an old disk (think 1970) works better if no alternate tracks are
assigned than if some are, for critical locations in the disk.
--
Cheers
John
-- spambait
[EMAIL PROTECTED] [EMAIL PROTECTED]
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list
