Performance in Intel IOMeter Database Pattern

The Database test opens up the show. This pattern reveals the drives capability to process a mixed stream of requests for reading and writing random-address 8KB data blocks. By changing the ratio of write and read requests we can estimate how well the drive is sorting the requests out.

The drives are grouped according to their manufacturer to make the table smaller:

For better readability, we draw diagrams for different workloads.

The drives from Hitachi behave much like each other under a small load (one request). When there are many write requests (more than 50%), the HDS 722512VLAT20 starts lagging behind the “older” mates. This situation lasts until 90% writes, when the drive “realizes” that there are so few read requests that it’s not necessary to allot much cache memory for them, but better reserve it for lazy writing.

We increase the workload to 16 requests.

The story repeats: Hitachi HDS722512VLAT20 is quite indifferent to the increase of the writes share until there are 100% of them in the queue. In this mode, however, it is the fastest of all. It seems like it has more segments in the buffer, although a small overall buffer size (compared to the IC35L120AVV207-1).

Now, the workload is maximum, 256 requests.

The drives are faster in this test and keep on their speeds throughout the test, save for the HDS722512VLAT20.