Database Pattern

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% throughout the test while the request queue size varies from 1 to 256.

I built diagrams for each of the drives.

The drive shows good scalability of performance depending on the request queue depth. The deferred write algorithms do their job here. There is nothing extraordinary in this diagram overall.

Easy to see, the graphs of these two drives are almost identical. It means that they both have identical HDDs, obviously with a Seagate logo, inside. Note that the right part of the graphs is almost flat, indicating a lower efficiency of the deferred write algorithms in comparison with the previous drive.

The Teac HD-15 PUK-B cannot boast efficient deferred writing, either. 

Note how different the left parts of the graphs are while the right parts are similar. It is an example of the reordering of read and write requests by means of NCQ technology and of the benefits it provides. Of course, this is not an ideal performance because there is but a small increase in the middle part of the diagram (when there is the same share of reads and writes in the queue), yet the effect of NCQ is visible.

The drives from Toshiba and Transcend seem to be based on the same HDD as the Teac HD-15 PUS: they have normal deferred writing and no reordering of read requests.

The firmware or the USB controller of the ZIV 2 is somewhat less efficient, but the overall shape of the graphs is normal.

Note the sudden slumps in the graphs which are strong with the USB interface and almost imperceptible with FireWire. This must be due to some flaws in the operation of the USB controller of that drive. The question is if these flaws show up in the synthetic test only or in real-life applications as well.