In the Database pattern the tested 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 this link to view the tabled results for the IOMeter: Database pattern.
We will build diagrams to illustrate the SSDs’ performance at request queue depths of 1, 16 and 256.
We’ve got interesting results at a queue depth of 1. The Intel is unrivalled at pure reading, followed by the Super Talent MasterDrive SX which is a little faster than the Indilinx-based models. The Super Talent MasterDrive SX then slows down at 20% writes, though. The Intel X25-M holds on to its leading position until 50% and more writes when it gives way to the Wintec. Interestingly, the other two SLC-memory drives are not that fast although have the same controller inside.
Kingston’s SSDs are all on the losing side here. The V+ series model is somewhat better than the others, but still no good for server applications if the server has any writing to do.
When the queue grows longer, the Intel X25-M is first, followed by the Indilinx-based models among which the Wintec stands out. The other SSDs seem to be only any good at reading.
Winding up this part of our tests, we will show you diagrams with each SSD’s performance at five different queue depths. Such diagrams are usually so characteristic that you can easily name the controller installed in the particular SSD.
Kingston’s SSDs all behave in the same manner. Judging by the diagrams, we can suspect the Toshiba controller to trace its origin back to the JMicron. Yes, it is faster than the latter but not fundamentally different. Interestingly, Toshiba’s controller just ignores a request queue if there is any.
The 128GB V series model is somewhat faster than its series mate here.
These graphs tell us a lot indeed as they are highly typical of SSDs with Samsung’s controller inside. When there is a request queue, the SSD slows down at processing a small percentage of writes but accelerates at pure random reading.
These four SSDs all draw one and the same graph. We can see a characteristic effect of the request queue, good processing of writes, and stable performance. They even all have a performance growth in one and the same point. The minor variations in the shape of the graphs are obviously due to the flash memory modules: the MLC model is somewhat faster at reading but worse at writing. The Wintec is a little slower than its opponents under high percentages of writes.