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: IOMeter Database table.
We will build diagrams for request queue depths of 1, 16 and 256.
Request reordering is impossible at the minimum queue depth. The speed of the heads and deferred writing are the only two factors a drive’s performance depends on here. The Blue and the Black models from Western Digital are still ahead of the others at reading, but give way to the Samsung and Seagate when write requests prevail. The Samsung is the faster drive in this pair: the Seagate only outperforms it at pure reading. The WD Caviar Green is surprisingly good. It is quite competitive at high percentages of reads. By the way, its graph has almost the same shape as the graphs of the other HDDs from WD. They seem to share the same firmware algorithms.
The queue is now 16 requests deep. Let’s see how the performance of the drives increases. The standings do not differ much: the HDDs from Western Digital are still in the lead but give way to their opponents at high percentages of writes. The Samsung is somewhat better than the Seagate at writing but worse at reading. The Seagate must be equipped with more efficient NCQ algorithms. The WD Caviar Green is almost as fast as the Samsung at reading. That’s an excellent performance for a HDD that has a lower spindle rotation speed than the others. Interestingly, the Green one has more effective deferred writing than the Blue and the Black models.
When the queue is even longer, we can see all the things we’ve seen above plus a few more facts. For example, the Caviar Black breaks away from the Caviar Blue at reading: its double processing power seems to come into effect. The Seagate is ahead of the Samsung at every load, including writing. This must be due to its larger cache buffer. The Samsung has minor flaws in its firmware for some reason that lead to a sudden performance hit at 10% writes.
To sum up this section of the review we’ll publish the drives’ graphs for five queue depths.
The Samsung behaves just like the other SpinPoint F1 series models we have seen before. NCQ is here and even more effective than in the firmware of Samsung’s earlier series. Deferred writing can be observed, too.
It is a most fascinating thing to watch Seagate’s firmware evolve. And as the recent story shows, sometimes you’d better watch this process from the outside. Here, we can see that the progress has reached a certain limit already. The drive has request reordering and good deferred writing. There are no problematic sections in the graph. But judging by the test results, this firmware can still be improved to the level of Western Digital’s products at least.
Western Digital seems to have developed a single basis for all its firmware. The company’s latest drives (even 2.5-inch ones) all come with firmware that shows the same behavior. And there is no problem in that: the firmware is good indeed, especially in terms of request reordering. Deferred writing might be better, though. The present implementation of this technique is not very effective.