Re: which dual-CPU hardware/OS is fastest for PostgreSQL?

...and on Mon, Jan 10, 2005 at 08:31:22PM -0800, Joshua D. Drake used the keyboard:
>
> >
> >RAID controllers tend to use i960 or StrongARM CPUs that run at speeds
> >that _aren't_ all that impressive.  With software RAID, you can take
> >advantage of the _enormous_ increases in the speed of the main CPU.
> >
> >I don't know so much about FreeBSD's handling of this, but on Linux,
> >there's pretty strong indication that _SOFTWARE_ RAID is faster than
> >hardware RAID.
> >
> >
> Unless something has changed though, you can't run raid 10
> with linux software raid and raid 5 sucks for heavy writes.
>
> J

Hello, Joshua.

Things have changed. :)

From 2.6.10's drivers/md/Kconfig:

config MD_RAID10
        tristate "RAID-10 (mirrored striping) mode (EXPERIMENTAL)"
        depends on BLK_DEV_MD && EXPERIMENTAL
        ---help---
          RAID-10 provides a combination of striping (RAID-0) and
          mirroring (RAID-1) with easier configuration and more flexable
          layout.
          Unlike RAID-0, but like RAID-1, RAID-10 requires all devices to
          be the same size (or atleast, only as much as the smallest device
          will be used).
          RAID-10 provides a variety of layouts that provide different levels
          of redundancy and performance.

          RAID-10 requires mdadm-1.7.0 or later, available at:

          ftp://ftp.kernel.org/pub/linux/utils/raid/mdadm/

There is a problem, however, that may render software RAID non-viable
though. According to one of my benchmarks, it makes up for an up to
10% increase in system time consumed under full loads, so if the original
poster's application is going to be CPU-bound, which might be the case,
as he is looking for a machine that's strong on the CPU side, that may
be the "too much" bit.

Of course, if Opteron is being chosen for the increase in the amount of
memory it can address, this is not the issue.

HTH,
--
    Grega Bremec
    gregab at p0f dot net