[BlueOnyx:07276] Re: Sata300 Disk Transfer Speeds

[Michael Stauber]

Hi David,

ST31000528AS:
hdparm -I /dev/sda | grep -i speed
           *    SATA-I signaling speed (1.5Gb/s)
           *    SATA-II signaling speed (3.0Gb/s)
hdparm -tT /dev/sda

/dev/sda:
 Timing cached reads:   29492 MB in  2.00 seconds = 14780.61 MB/sec
 Timing buffered disk reads:  346 MB in  3.01 seconds = 114.99 MB/sec

------

ST31000528AS:
hdparm -I /dev/sda | grep -i speed
           *    SATA-I signaling speed (1.5Gb/s)
           *    SATA-II signaling speed (3.0Gb/s)
hdparm -tT /dev/sda

/dev/sda:
 Timing cached reads:   6296 MB in  2.00 seconds = 3155.00 MB/sec
 Timing buffered disk reads:  306 MB in  3.01 seconds = 101.62 MB/sec

------

Hitachi HDS721010CLA332:
hdparm -I /dev/sda | grep -i speed
           *    SATA-I signaling speed (1.5Gb/s)
           *    SATA-II signaling speed (3.0Gb/s)
hdparm -tT /dev/sda

/dev/sda:
 Timing cached reads:   34140 MB in  1.99 seconds = 17127.48 MB/sec
 Timing buffered disk reads:  390 MB in  3.01 seconds = 129.63 MB/sec

------

However, all these tests were run on boxes with 6-12 VPS's running, so that's 
just a momentarily snapshot at the currently available I/O bandwidth that 
doesn't show the maximum possible speed.

All of these boxes use the onboard SATA controller. Box #1 and #3 are using 
Asus boards of a kind which you'd rather find in desktop PCs. #2 is also using 
consumer grade desktop equipment, but I forgot what kind of board is in it.

----

Here is some output from two further boxes, which both use identical server 
grade hardware and are almost bare of any load at the moment:

hdparm -tT /dev/sda

/dev/sda:
 Timing cached reads:   23976 MB in  1.99 seconds = 12048.71 MB/sec
 Timing buffered disk reads:  706 MB in  3.01 seconds = 234.18 MB/sec

/dev/sda:
 Timing cached reads:   20728 MB in  1.99 seconds = 10409.76 MB/sec
 Timing buffered disk reads:  638 MB in  3.03 seconds = 210.63 MB/sec

The disks are 1.5TB WD's or Seagates, but these boxes use server boards with 
dedicated add-on RAID controllers, which may explain why the buffered reads 
are almost twice as good as on the boxes further above, which just use the 
rinky-dinky onboard controllers.

-- 
With best regards

Michael Stauber