[PLUG] Partitioning Schemes

[Christopher E. Brown]

On Tue, 16 Jul 2002, Steve Bonds wrote:

> On Tue, 16 Jul 2002, Felix Lee flee at aracnet.com XXXXXXXXXXXXXXXXXXX wrote:
>
> > I think, position of partitions on a disk is generally pointless
> > micro-optimization.  My instinct is if there is a difference, it's not
> > a big enough difference to matter for average use.  Benchmark it and
> > make sure.  This type of thing needs to be benchmarked; it's quite
> > hard to predict the right answer, especially now that disks are very
> > smart devices.
>
> I'll definately agree with this.  However, I understand that some folks
> have lots of free time and want to play around.  ;-)
>
> Some drives' (and SCSI cards') translation from SCSI blocks to disk blocks
> do not necessarily follow the start-at-the-outside-and-work-your-way-in
> convention.  This is especially true if using CHS addressing rather than
> LBA.
>
> Benchmark to be sure, but if performance is that critical, then there are
> lots of better things to do.  (Caching, striping, etc.)
>
> Be careful of developing some bizarre, hard to maintain disk layout in the
> name of performance.
>
>   -- Steve



Yes, but for the most part it holds true.  By habit I locate my swap
at the beginning of the disk, it helps.  For most other things it
really does not matter though.  It can easily be a 15% difference in
throughput inner to outer, but only assuming you are not bottle necking
elsewhere.  Other than when you are heavily into swap it makes little
difference.



-- 
I route, therefore you are.