On Tue, Jun 22, 1999 at 05:17:33PM +1000, John Leach wrote:
> Hi.
> In the Software Raid HOWTO the statistics indicate that Raid-10 while
> faster than Raid-5 is slower than Raid-0. I'm surprised that this is the
> case - does anyone have any other experience with Raid 10 ? If you have
> the disk space it sounds like a nice fast option but with redundancy.
> I'm keen to try it out if I can borrow a 4-disk system, I'll let the
> list know...
I think most of the performance degradation I saw with RAID-10, was due
to memory bandwidth problems in my machine. There's no way for me to get
more than 33 MB/s (or so) from four disks that each do 12 MB/s.
RAID-10 must copy the same information (RAID-1) to the stripes (RAID-0),
so there's obviously a lot of information copying going on.
Perhaps the new ext2/vfs/buffer/cache changes in 2.3.6 will change this.
>From what I could understand, the kernel does a lot of unnecessary copying
of pages between buffer/cache/etc., and these unnecessary copies have been
eliminated in the latest 2.3 kernel. I haven't tested it yet though, because
Linus wrote that he released the kernel an hour after he fixed the last bug
that ``seemed'' to cause fs corruption, and my benchmarking-box is also my
production box :)
But if anyone feels like playing around with very experimental kernels, I'm
sure a lot of people here (and on linux-kernel) would love to hear about the
results.
Cheers,
................................................................
: [EMAIL PROTECTED] : And I see the elder races, :
:.........................: putrid forms of man :
: Jakob Østergaard : See him rise and claim the earth, :
: OZ9ABN : his downfall is at hand. :
:.........................:............{Konkhra}...............:
