On Thu, Oct 31, 2002 at 11:47:03AM -0800, Chris Wedgwood wrote:
> On Thu, Oct 31, 2002 at 08:42:33PM +0100, Ragnar Kj?rstad wrote:
> > If you do the same calculations when multiple operations are going
> > on at the same time you'll find that if you're reading 10 files from
> > 10 disks it's going to be a lot quicker if the files are _not_
> > spread out.
> More disks -> more heads -> lower average seek time.
Obviously more heads is better, but spreading each individual file over
multiple disks is not always faster - in particular when there are many
requests executed at the same time.
> If you experiment with this, you'll find eads over multiple disks and
> stripping is almost *always* faster. RAID4/RAID5 will have slower
> writes obviously though.
Try the following:
Run a benchmark with 100 threads (or more) on a system with e.g. 10
drives. You will find that concatenating the drives is faster than using
RAID0. All drives will be busy all the time in both setups, but with
RAID0 you will have 10 seeks pr file instead of one, so total
performance will be significantly lower.
There are a number of external factors, like readahead, caching,
blockdevice-elevator and so on that affect the result, so it's no easy
to predict how big the difference will be.
If the filesystem uses different layouts depending on the low-level
storage system that will affect the result as well. How does XFS handle
underlaying RAID0 compared to concatenated devices?