Girderman

justpassingby & I have gotten into a side-discussion on whether or not partitioning a HD and allocating certain files (the O/S) to one partition and everything else to the other (I think that's the situation) will optimize the access to the drive.

Full context Here

He says it does and I have my doubts. My operating assumption is that the hard drive itself is the rate determining step in the process, and so whatever is done on the platters is going to produce neglible results.

Basically, my understanding is that the only real "optimization" occurs with the pagefile, and only then if it's on a completely separate physical drive. This is what I have read.

Who's right ? And if justpassingby is correct, how "right" is he ? Is the optimization significant ? Or neglible ?

Thanks in advance,

Girdie

justpassingby

And in he left corner, the challenger
Good idea you had there to start a new thread Girderman. Hopefully we'll reach to some definitive conclusions here as I'm wanting to learn the truth as much as you are.

Just wanted to quickly jump in and add a few remarks :

I don't think you'll change a turtle in a hare only by optimizing the partition's layout. There are lots of things that need to be done beforehand.

The starting point would have to be a clean and already optimized to the max computer. My point is I'm just trying to grab out some more milliseconds in here. My system being fast and responsive surely is more related to not having crapware installed than on the partition layout I'm using.

My system is on the first 15GB partition of a WD raptor 150GB drive. That partition has all windows files, microsoft office (custom install), antivirus, and all programs I use everyday or I want on that partition for some particular reason (adobe, vlc, winrar, imgburn, firefox, quicktime alternative, XXclone, ...).
Second partition on the raptor is a 45GB partition where I install bigger programs that are not related to windows and which files probably aren't needed in the boot process. It contains mostly games. I could have used a smaller partition for that one since even in my worst nightmares I have never used more than 25GB for my program files (mostly from having 5 games installed at the same time).
Third partition (rest of the space) is for downloads and medias files.

The swap file is on the first 5GB partition of my second Hitachi Desktar 320GB drive. It's a 2GB fixed size file. Rest of that drive is used for storage, backups and ghosts.

I want to specify that this layout is adapted to my own specific needs. I don't do video editing nor use 3D studio or Photoshop. All I want with this scheme is to have my system and boot files localized on the fastest part of the drive, without any other files interfering, and to ensure all future system updates will stay close of the current system files. Fragmentation is not much of a problem since most of my downloads will be done on a separate partition. And the swap file has its own disk.

justpassingby

I totally agree on the first part : you'll never have better performances on a 5400rpm disk than on a 10.000rpm Raptor. That's why laptops are horrible when it comes to drive's performances.

But I believe optimizing the position of the files on the platter can produce better results on identical drives.

The theoretical basis of my argument is that the transfer rate is better on the outer border of the platter than it is in the center, and from what I understood, it would be logical to assume that keeping the files that are actually used by the OS close to each others would reduce the time lost for head positioning.

Now I have the same interrogation as Girderman : will the gains be negligible or could we actually gain some seconds on startup or when running applications doing this ?

Now the idea of keeping one partition for the OS still has the benefits of reducing the fragmentation factor of that partition. But again that's more related to the habits of the user. If you can't refrain from downloading/storing your stuff on your desktop or in your documents folders, this layout is not for you.

Kalim

Well, the point of discussion for me is the concepts discussed herein and I've come across this mentioned before a few times already.

It is better, yes. When the system searches for data it has less data blocks to trawl through and search in order to find it or write to it. This is when read and write operations are executed. Obviously also in system wide searches, drive scans, synthetic benchmarks, database recollection, large program loadup (etc). Thus it's quicker and it preserves the HDD integrity in a more satisfactory way. You will experience better results like this, hence why Linux builds use separate partitions for page file and most Linux users will keep one partition for OS and another for data too. In fact, most corporations that I've come across make practice of the same habit. I know ours does, and it's one affiliated with the mainstream government who I know also practices the same in their hierarchical departments for better HDD/data integrity and access. Each partition can have a separate page file, which also adds a little boost.

Having separate HDDs is teh main deal though and much better as more simultaneous data can be accessed, hence why RAID striping is best for performance.

Personally, I am using this concept on my work systems and on 2 of my home ones (those with sufficient space). But mine is a little more customized; OS is HDD C, programs are HDD D, storage/documents are HDD E and backup 1 and 2 which are external USB 2.0 and FW800 drives.

But in the end, it's a something dependent on variable endless factors. It's more a matter of 'marginal' speed gains, better file integrity, better practice and organization than anything else. IDK to what extent, but enough for it to be common convention in places where it matters, hence I tend to see it as a wise move if one feels a little challenging. Nothing I would consider a need to go too deep into though.

Girderman

Okay, so the "friend" back in the other thread, that wanted to partition his friend's HD to optimize it wan't completely out to lunch, then.

Kalim

Just something I remembered [late] that's useful here (this is what happens when you don't study IT knowledge for years and tend to make little use of it in daily life to recollect, equipped with potentially no time, a completely unrelated profession and a heck busy life. ).

Data recovery is much much more advantaged for drives that are partitioned and allocated separate partitions for programs/windows/swap file/data, and thus in effect data loss is by far more minimal. A partition physically dissects a HDD into separate parts. Why it's advantaged is, because as you bootup/shutdown your system (minimal practice) there are many small files written to the HDD. If you have utilized good practices and well planned disk management, the bootup will be involve a totally different partition than where the data is, thus the data drive partition is not affected nor written/read from/to, and any files that are deleted/lost that you may need to recover, can still be accessed pretty easily, with a very high chance of full recovery (you'll see data recovery firms mention this in essays). Data loss is also more minimal based on this principle where the HDD corruption of a few program/temporary/swap/windows files does not affect your data partition or files.

This also goes for general system scans (especially deep), bootup speed and HDD performance (by far, believe me, mine has corrupted to SMART status 'Good' from 'Excellent' during overclocking in the past week and the bootup now takes 3 minutes 20 seconds, up from 15 seconds!). If one partition is corrupting, another (unless physical) will not be affected and neither the access to it or it's performance degraded.

Same principle with fragmentations and physical drive errors.