In the Database pattern the drive is processing a stream of requests to read and write 8KB random-address data blocks. The ratio of read to write requests is changing from 0% to 100% with a step of 10% throughout the test while the request queue depth varies from 1 to 256.
You can click the following link to view the tabled results in IOMeter: Database.
We will build diagrams for request queue depths of 1, 16 and 256. There are too many results, so we divide them into two diagrams here and in some of the later tests.
Alas, our apprehensions come true: the Seagate does not have deferred writing at all. And it is not really fast at reading, either. It is interesting to compare the leaders: the Western Digital accelerates steadily as the percentage of writes in the queue is increasing whereas the Fujitsu wakes up rather late but speeds up fast, outperforming the Western Digital at pure writing. Anyway, we like the behavior of the HDD from Western Digital more because it has the best results under the most complex loads (when the hard disk is receiving about the same share of reads and writes).
Request reordering begins to work when we increase the load. This saves the day for the Seagate when the percentage of reads is high (this HDD has efficient NCQ) but it has no deferred writing anyway. The HDD from Western Digital is the obvious leader still. The Fujitsu might challenge it, but had a sudden slump at 10% writes. Its firmware is not ideal, either.
When the request queue is very long, the driver begins to affect the performance. Thanks to it, the Seagate speeds up at low percentages of writes and even shows something like deferred writing. It is far slower than the other HDDs, though. The Western Digital is brilliant again while the Fujitsu is second best.
And now for each of the three new drives we will publish a diagram with five graphs for five request queue depths. This helps better understand the behavior of a HDD and the specifics of its firmware algorithms.
Comparing this diagram with the one of Fujitsu’s previous series, MHY 2BH, we can see the progress in this company’s firmware. The efficient deferred writing is now combined with efficient request reordering algorithms. If it were not for the slumps at 10% writes, the firmware would be perfect.
The Hitachi drive is consistent under any load, showing good scalability and medium-efficiency deferred writing. It is good that there are no slumps in the middle of the graphs, but NCQ might be more effective.
We have said all we think about the Seagate drive above, so we can only repeat our point. This model is very slow at reading and does not support deferred writing. Its efficient NCQ algorithms cannot help it much.
You can view such diagrams for the other participated HDDs in the above-mentioned reviews.