Articles: Storage

Bookmark and Share

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 ]

Multithreaded Read & Write Patterns

The multithreaded tests simulate a situation when there are one to four clients accessing the hard disk at the same time – the clients’ address zones do not overlap. We will discuss diagrams for a request queue of 1 as the most illustrative ones. When the queue is 2 or more requests long, the speed doesn’t depend much on the number of applications. You can also click the following links for the full results:

The Hitachi drive with newer firmware delivers a fantastic result. We would praise it loudly if it were not for one thing. This HDD is as fast as 139 MBps at four data threads, which looks improbable. And if you take a look into the table, you will see amazing data-transfer rates like 160 and 170 MBps. So, this is obviously an error of the benchmark. Even with a narrow test zone and highly optimized firmware, the HDD has to suffer a performance hit, however small it may be.

The considerable fluctuations of speed at different request queue depths are hard to explain, too. We have read through all available results carefully and found nearly all of Hitachi’s latest-generation HDDs with updated firmware to behave like that. The Seagate Constellation ES, according to the table, also manages to be faster at multiple threads than at one thread. The distribution of speed among the threads does not provoke any suspicions, by the way. Every thread is being read at about the same rate. Summing it up, we have to be skeptical about these results and are trying to find the cause of this problem and ways to solve it.

As for the more credible results, the pair of WD’s latest-generation HDDs (with two platters inside) are somewhat faster than the others. The Samsung EcoGreen F3 shows the typical behavior of this company’s recent models: it copes well with two and three data threads but slows down to 10 MBps at four threads.

There are no extraordinary results at multithreaded writing except that the leading Seagate Constellation ES looks suspicious at long request queue depths, but we can explain this by its well-developed caching. Its performance at a request queue depth of 1 is excellent, but within reasonable limits. It is the power-efficient HDDs and the Hitachi (with both firmware versions) that suffer the biggest performance hit from having to write multiple data threads.

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 ]


Comments currently: 6
Discussion started: 04/19/10 07:58:21 PM
Latest comment: 10/27/11 05:48:36 AM

View comments

Add your Comment