Database Patterns

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 for IOMeter: Database pattern.

We will build diagrams for request queue depths of 1, 16 and 256.

When the queue is only 1 request deep, an HDD has nothing to reorder. It can only process read requests as fast as it can and put write requests into its cache. The 500GB model from Western Digital is superior at both and enjoys a nice lead over its 320GB cousin. The Samsung has less effective deferred writing algorithms: its performance does not grow up much as the percentage of writes increases. The Hitachi is different: it obviously has very effective deferred writing algorithms but it does not like mixed loads (with about the same share of reads and writes).

NCQ algorithms come into play at a queue depth of 16 requests. Western Digital’s drives feel all right and both behave in a similar way. Judging by the performance growth in the left part of the diagram, these HDDs have NCQ. The Samsung still shows poor deferred writing and has no request reordering. The Hitachi is more effective than the Samsung at pure reading and writing but is very poor under mixed loads. So, it is hard to tell which one is the worse of the two.

There is nothing exciting when the queue depth is increased further. Western Digital is still in the lead. Interestingly, the lower-density 320GB model is better at high percentages of reads – it seems to have more effective NCQ algorithms. The 500GB model is better at processing write requests, though.

The other two HDDs are not quite good: the Samsung prefers to process requests as they are, without any optimizations, and only applies some deferred writing to improve performance. The Hitachi shows excellent deferred writing and good NCQ algorithms but slows down under mixed loads.

The following diagrams show the performance of the HDDs with 250GB platters at five request queue depths.

Indeed, the Samsung has no NCQ at all. Its performance is nothing but mediocre in this test. Its 320GB cousin was no record-breaker, either, but was faster in server tests and had some kind of NCQ.

Western Digital’s firmware for 2.5-inch drives is nearly ideal. The 500GB model has inherited excellent NCQ algorithms and has even improved somewhat in terms of deferred writing.

You can find such diagrams for the other two HDDs in the above-mentioned reviews.