Articles: Storage
 

Bookmark and Share

(6) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 ]

Database Pattern

In the Database pattern the HDD 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.

The following diagrams show the dependence of speed on the ratio of reads to writes.

There is no fundamental difference between the firmware of the two models. The number of operations per second is determined by the spindle rotation speed at every ratio of reads to writes. Overall, the HDDs show the typical behavior of modern 2.5” drives with a slump in the middle of the diagram: reads and writes in equal proportion is the most difficult case for firmware which cannot decide if the buffer segments should be allotted for deferred writing or for look-ahead reading.

The Hitachi shows good scalability of performance depending on load. Its deferred writing strategy is very effective as is indicated by the rise of the graphs in the right part of the diagram. Take note that there is almost no slump in the middle of the diagram as opposed to the graphs of the Fujitsu HDDs.

Samsung acts up again, drawing flat graphs with odd zigzags on the right. This firmware may be defective or, like with the 160GB model, this is the result of incompatibility of the firmware with this particular test. The drive’s real-life performance may be better.

The WD2500BEVS is almost perfect in this test. There are no slumps in the graphs. The drive’s performance increases in proportion to the share of writes at every request queue depth. That’s an example of a good implementation of deferred writing.

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 ]

Discussion

Comments currently: 6
Discussion started: 01/15/08 07:15:04 PM
Latest comment: 04/30/08 05:09:53 PM

View comments

Add your Comment