Interesting hard drive testing

[Bozo]

http://www.xbitlabs.com/storage/defrag-tests/


Bozo

 

[Buck]

Defragmenting is a good way to exercise a drive and cause it to go through a series of short and long seeks.

 

[Mercutio]

Defragging seems to me to be a very good real world test of driver performance. I wonder why I haven't seen it done before.

 

[Tea]

Strikes me that the really interesting matter raised is the way that different defrag packages have such different results on different drives. I would have though that the operation of defragging would involve much the same sort of low-level disc operations no matter which program you were using - i.e., that any given drive would be a "good defragger" or a "poor defragger" and be fairly consistent over different packages. This is not the case, it seems. Hmmm...

 

[LiamC]

I would have thought that this is a next to useless "benchmark" because there is no repeatability on where the files are placed physically on the hard drive. So one run might be with a less fragmented file system over another - and there is no way to know before hand or after, how this might affect the workings of the software.

You might be able to make some assumptions after doing may (>30) runs for each drive with the same file set, and some statistical analysis, but I didn't see any in the article...

 

[Mercutio]

The article seems to indicate that the backup/restore practice they used fragmented the files in a predictable manner. So at least for the purpose of their test, there was repeatability.

 

[LiamC]

I'd be cautious about accepting such as statement

The article seems to indicate that the backup/restore practice they used fragmented the files in a predictable manner

Random chance says that the file level fragmentation could favour one particular drives file placement over another. IMHO, there are two few iterations of the test to be sure, but it could be verified.

It's an interesting idea, but they haven't done enough work to prove the concept. I'm happy to be proved wrong of course...

 

[Cliptin]

I am pretty surprised that the fragmentation level changed after the imaging.

 

[Cliptin]

While this article is about hard drive testing, the data can be used to evaluate defrag software. OO defrag took,usually, more than twice as long to complete than the fastest contender. While there is no data indicating quality, speed is half the equation.

 

[Buck]

Part of the speed of defragmenting has to do with the swap file. If the swap file size is dynamically controlled by Windows, the process time is again different than if the swap file size was pre-determined. Additionally, the location of the swap file alters this process since seeks would either be longer or shorter. If all of the circumstances regarding fragmented files and system files could be defined, then degramentation could be used as a reliable benchmark.