Disk Response Time
For 10 minutes IOMeter is sending a stream of requests to read and write 512-byte data blocks with a request queue of 1. The total of requests processed by each HDD is much larger than its cache, so we get a sustained response time that doesn’t depend on the HDD’s buffer size.
It is very hard to improve the read access time of HDDs due to their very operating principle. As usual, the WD Caviar Black series (and their server cousins) have no rivals in this test as they have the fastest read/write heads today. Of course, they produce more noise when moving their heads around, yet that noise is very far from the awful ramble we used to put up with a few years ago. As for their response time, it is as good as can be possibly achieved from 7200RPM hard disk drives. Interestingly, the 1.5-terabyte models are faster than their smaller-capacity cousins. The WD Caviar Blue is somewhat slower, yet even this drive is better than the Hitachi team which, in their turn, are faster than the 7200RPM Samsung.
When it comes to HDDs with lower spindle rotation speed, the 1.5-terabyte Samsung EcoGreen F3 is better than the others. The old WD Caviar Green and the Samsung EcoGreen F2 are the only low-speed drives that have downright poor results.
The two 1.5-terabyte models from Western Digital are unrivalled at writing. They make the most from their advantage in the way of 64 megabytes of cache memory.
It’s interesting to compare the HDDs with 4KB sectors: Samsung seem to have developed a better algorithm of writing 512-byte data blocks because their HDD is much more responsive than the Western Digital products with Advanced Format.
Random Read & Write Patterns
Now we will see how the performance of the drives in random read and write modes depends on the size of the requested data block.
Everything goes as it should when the HDDs are reading random-address data blocks. Their speeds are determined by their response time.
Take note that it’s only with 64KB and larger data blocks that we can observe a notable decline in the amount of operations per second. As you can guess, it is when the time spent to read the data block becomes an important factor. With smaller data blocks, the performance depends on the time spent to find the requested block while the read operation itself takes but very little time.
It’s different at random writing. First, the HDDs with 64 megabytes of cache enjoy a huge advantage. A large buffer improves deferred writing, which has a positive effect on random-address write operations. Second, we can see the HDDs with 4KB sectors having a hard time when they have to write small data blocks. They have to do extra read operations and change whole 4KB sectors, so their performance lowers by half. It’s also interesting to check out the performance of such HDDs under unaligned load, which is equivalent to writing to an unaligned disk: the HDD suffers a penalty even with large data blocks because it has to write some part of the data in the small-data-block mode, i.e. by reading the data block first.
But what do we have with the WD Caviar Green S8? That model didn’t behave like that in our earlier tests. Our Caviar Green S8 seems to have some problems with deferred writing. Well, it is going to be interesting to see how the lack of that technology will affect the drive’s performance.