To complete our discussion about the WD2500JD and WD360GD drives let's
test the RAID1 arrays based on these drives. Actually these tests
are of just cognitive interest as I can hardly imagine a person who could
decide to mirror such discs. Just because there's no practical use of it
I tested only the SATA drives as both the interface and the controller
are relatively new.
As expected, the results haven't changed. The linear read diagrams can be found
here.
Like in case of the distributed array, the access time of the WD360GD is a
bit better.
The Disk WinMarks look worse for the mirrors - there must be no cheats
in mirroring at all.
Intel IOMeter
Mirroring doesn't affect scores of the WD2500JB, but in case of the
WD360GD the array's performance looks better than that of the single drive
as the load grows up. The matter is that data can be read from any of two
array discs. And such effect results from the small volume and high speed
of rotation.
This diagram proves it.
The WD360GD has quite strange scores in the random writing test - the
character of the diagrams looks too different.
In the Database patterns with various relations of read/write requests
the array of two WD360GD looks completely different from the single disc.
The advantage of mirroring is that data are read from a more appropriate
disc.
Now come the serial modes. The disc gets requests for reading or writing
of data blocks the size of which gradually grows up. The queue depth equals
4 (very light).
The single WD2500JD and the array have very close scores, in contrast
to the WD360GD.
The scores for the mirrors of the WD360GD and WD2500JD look similar
at writing.
Conclusion
The mirror of two WD2500JD performs similarly to the single disc (except
for the slower serial writing). The array built on the WD360GD behaves
differently both at writing and reading.
