nano wrote:
> Hi Paul
>
> Thanks for writing. I backup all of my data files regularly to external
> hdd a couple times a week and online several times a day. Backups are
> good! The raid consideration is to guard against a single disk going
> out, to catch the interim updates. Also, if one disk starts to go out,
> the system is not at risk, I'd need to troubleshoot or replace the
> failing disk.
>
> Still curious about a couple of things.
>
> #1 raid 1 would likely be the best variant for my simple needs?
> #2 is the p5b deluxe onboard raid controller ok, or should I buy a
> controller?
> #3 if I replace the raptor and my os/boot drive is a larger,
> partitioned, raid 1 set, will performance suffer?
> #4 not really sure about boot disk issue in general. If only two disks
> in the box, if software raid, I think I need a c partition that is not
> raid, to boot from, and other partitions would be in raid array?
>
> I took the faltering maxtor out of the case and have it laying on it's
> side next to the case; it's gone a day without issues. It couldn't be
> overheating issue in case becase it has failed after long off periods;
> cabling maybe, though I've changed them.
One thing that bothers me about RAID, is the general inability to tell
when things aren't working right. Yes, a bad status on a drive, such
as a failure detected by the controller on the disk, would work OK.
But I'm thinking of more subtle failures.
To give an example, someone had a SIL3112 based RAID mirror they set up.
One day, one of the disks failed, leaving one working disk. The owner
of the system, discovered that the mirror disk, had stopped updating
like three months earlier, so in fact it wasn't a mirror. And apparently
that person received no warning that they could remember (even though,
I would have expected, the RAID BIOS should have complained on every
reboot, that something wasn't right - the status should have indicated
degradation for those three months).
In principle, what I'd want to see in a RAID1 design, is the ability to
check that the two disks are exact mirrors (offline check, like in the
BIOS, before the system boots). I realize, that commanding
the array to do a rebuild, guarantees that the disks are identical. But
it also runs the risk, of copying a "stale" image (like the three month
old example above) to the second disk.
1) Yes, RAID1 is an effective option. It is simple to understand, unlike
facing a broken RAID0+1 or RAID10.
2) The Southbridge RAID should work fine. Performance is good, as there
is a good connection bus-wise, to the rest of the system.
3) A RAID1 has the opportunity to accept read data from the first of
the two disks that are ready. In terms of raw bandwidth, there are now
some drives, which offer slightly more bandwidth than the Raptor. Where
the Raptor would excel, is in seek time. So on random access, the Raptor
might still be faster (bootup might still be faster on the Raptor).
4) You can place the boot volume on the RAID1. The RAID BIOS module in the
BIOS flash code, takes care of reading from the RAID, during bootup.
So that isn't a problem.
What is a problem, is getting a RAID driver onto the thing. If you ever
wanted to use RAID for booting, you'd enable RAID in the BIOS, connect
one disk, do F6 and install a RAID driver, even though there was only
one disk present. The Intel software supports migration, so you can go
from a single disk, to a RAID1 or a RAID0 or other form of RAID.
To take an OS install, in which no RAID driver was installed, and
move it to a RAID array, needs to get the driver installed somehow.
If you enable RAID, you cannot boot the existing disk. If you leave
the Southbridge in non-RAID mode, you cannot install a RAID driver.
So that is one of the challenges you'd face, if trying to put the
boot volume on the RAID. You need to be "RAID ready" from day one.
http://download.intel.com/support/ch...startguide.pdf
Paul