Mailinglist Archive: opensuse (1495 mails)

["David C. Rankin" <drankinatty@xxxxxxxxxxxxxxxxxx>]

Listmates,

        Looking at hdparm -tT results for a Linux Software raid 1 array, I see 
some
differences in the read performance that I don't know whether it is just
individual drive difference or whether mdraid may have something to do with it.
The mdraid 1 setup is on a pair of 750G Seagate drives:

/dev/sdb: ST3750330AS: 34°C
/dev/sdc: ST3750330AS: 36°C

        The raid1 devices are /dev/sdb1 and /dev/sdc1 which take up the entire 
disk on
each. I don't know if there is a reason not to use the whole drive in the
array, but the array is basically a temporary setup on my sons computer (with
open SATA port) to throw the drives in to get a week or so of runtime on them
before moving them over to a server.

        Checking read performance with hdparm of each individual drive and then 
of the
array itself shows a small performance difference/drop for /dev/sdc:

02:11 KillerZ~> for i in sdb sdc md0; do sudo hdparm -tT /dev/$i; done

/dev/sdb:
 Timing cached reads:   2130 MB in  2.00 seconds = 1064.71 MB/sec
 Timing buffered disk reads:  328 MB in  3.01 seconds = 108.94 MB/sec

/dev/sdc:
 Timing cached reads:   2088 MB in  2.00 seconds = 1044.08 MB/sec
 Timing buffered disk reads:  292 MB in  3.02 seconds =  96.84 MB/sec

/dev/md0:
 Timing cached reads:   2116 MB in  2.00 seconds = 1057.94 MB/sec
 Timing buffered disk reads:  322 MB in  3.01 seconds = 107.13 MB/sec

        Thinking about it (dangerous thing to do) the questions arose of (1) 
how does
mdraid actually work, is there a primary drive that is uses for read/write
while the other is there for sync/backup?

        (2) If you can't write simultaneously to both at the same time using 
the same
clock cycle (can you??) so somewhere it seems there must be a write/copy
function in the mdraid scheme, and if the drives are slaved together by my the
data in mdadm.conf, could the read performance difference being seen with
/dev/sdc above be the result of mdraid function and not simply a drive variance?

        I haven't a clue, that's why I'm asking. And then there is the /dev/md0 
hdparm
performance which looks like it takes the better of the two performance stats,
less a tiny amount of mdraid overhead. How does that work? Or is the difference
there more of a simple weighted averaging of the drive performance between sdb
and sdc??

        Any software raid gurus willing to shed a bit of light on this 
situation.

-- 
David C. Rankin, J.D.,P.E.
Rankin Law Firm, PLLC
510 Ochiltree Street
Nacogdoches, Texas 75961
Telephone: (936) 715-9333
Facsimile: (936) 715-9339
www.rankinlawfirm.com
-- 
To unsubscribe, e-mail: opensuse+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: opensuse+help@xxxxxxxxxxxx