Disk optimization strategies and file placement

[Keatah]

According to disktrix, you're supposed to put directories and startup files at the very outer edge of the disk platters. Then your most frequently used applications. Then on the inner part of the disk, all your little used stuff like pictures and mp3's and zip archives.

Is this really the best strategy? Or can we do better?
Certainly you won't see 400% improvement gains like they claim, unless you have a very bad layout that's severely fragmented to begin with.

Considering modern drives with multiple platters, is it best, perhaps, to just be sure the files are defragmented? And let the disk firmware and OS determine placement?

Or perhaps it would be best to put the directory smack dab in the middle of all your frequently accessed data?

And finally, is the outer edge the very fastest? Especially considering that modern drives will either park the heads totally off the platters, or, let the heads float in the middle of the disk and then park them.

And, like I just said, is the outer edge the fastest? There's extra processing time done by the controller to work there, because, ~1GHz is the current limitation. Most recent drives top out at 1GHz, so data density isn't the highest at outer edge, it's somewhere in the last 3rd of the drive.

All in all, what's the strategy that's worked best for you?

[Cudni]

Originally Posted by Keatah

All in all, what's the strategy that's worked best for you?

For me, stopped bothering with it all and leave whatever optimiser running at its default settings.

[kupo]

Originally Posted by Cudni

For me, stopped bothering with it all and leave whatever optimiser running at its default settings.

Agreed

[Isso]

As a former developer of UltimateDefrag I can confirm that Disktrix strategy does make sense. The main reason is that modern HDDs are about 2 times faster on the outer edge compared to the inner.
Just test any HDD and you'll see that yourself, here's one example:
http://i84.photobucket.com/albums/k7...01FALS-00J.png

And it doesn't matter how many platters the HDDs have - pretty much any HDD has a similar behaviour. So for above mentioned HDD you'll receive 100% increase in linear reading speed if you move a file from the most inner track to most outer. This assumes that HDD has only one partition taking entire disk span.

Real-life improvement will be less, because random access is involved that is significantly reducing the I/O speed. To improve random access speed Disktrix proposes to split the files into two groups - High Performance and Archive and place them to fastest and slowest areas of the disk. This results in High Performance data to be more "packed" together, thus the HDD heads need to travel less distance to access them, improving the random access speed.
For example - an ordinary HDD head takes 15 ms to travel from most outer track to most inner. So if we have High Performance files randomly distributed across the HDD then the heads may take up to 15 ms every time they need to access a file located at the opposite track. But if we "group" all High Performance files together and place them in first 20% of the space, the head on average will take 5 times less time to move from one file to another.

One thing to consider is the internal cache of the HDD as well as OS file system cache. These are trying to cope with non-optimized file placement and significantly improve the read/write speed. Thus 400% improvement is of course just a theory. However on an average drive with moderate fragmentation one can expect a 20-30% of increase in speed after optimization.

If HDD is less than 30% full, or if the partition that you are optimizing is much smaller than total HDD size, then the file placement strategy doesn't make much sense. Defragmentation however is still needed if the file system is too fragmented.

I recommend to have a look at UltimateDefrag manual http://www.disktrix.com/downloads/Ul...3UserGuide.pdf, it's an interesting reading that explains all these in details and with nice illustrations.