>Jay West wrote:
> My primary reason for using raid1 is speed. However, the failover is a great
> added benefit (as long as the controller supports booting off the raid set -
> some do, some don't).
Jerome Fine replies:
I must be very confused and missing something as well. Please
set me straight.
First, there are two methods of RAID 1 (Mirroring) implementation:
(a) Using software within the operating system to perform each WRITE
twice and perhaps using software again to allow the operating system to
select the drive which has the head closer to a READ request. While
the latter will save some time, the former is probably a time waster (in
respect to a single drive system that does not support RAID 1). I
understand that Windows NT (Yeck!) uses this method.
(b) Use firmware within a separate controller and HIDE the second
drive from the operating system. (I have this type of controller, but
I found the RAID 1 firmware useless.) However, because my controller
is able to support UDMA 100 I/O requests (there is a Pentium III 750
and motherboard which supports ONLY IDE speed access), the
sustained throughput via the PCI controller for the EIDE hard drives
is perhaps FOUR times the throughput for these same hard drives
using just the motherboard. BUT, I don't believe it is fair to say that
the RAID 1 function (if it did work) is faster because the RAID 1
controller supports faster throughput. Indeed, since I rejected
the (totally incompetent implementation of the) RAID 1 firmware
and ONLY keep this PCI controller because the additional speed
is a big advantage, I tend to reject any conclusion that the RAID 1
function is faster in and of itself - it is the PCI controller that is
faster and if the RAID 1 function worked well, that would be a bonus.
So maybe we are each saying exactly the same thing, but from
a slightly different point of view - or perhaps Jay did not expand
what was said sufficiently to bring out the point that his RAID 1
controller is faster as well, but since the RAID 1 function does
work, both aspects have been joined into a single improvement.
As for backups, I just use the additional drives that I was going
to use in the RAID 1 configuration as storage for my daily
backup which takes only 10 minutes due to the speed of
the PCI controller for the EIDE drives. And while it does
take my personal time and attention, I also have the advantage
of having a whole layer of backups which I would not have
with only a RAID 1 configuration. This latter aspect is, in
my opinion, probably more of and advantage than having
only a single working backup at any time which is identical
to my primary copy from a hardware point of view, but could
contain corrupted information due to software problems.
I realize that each person will have a different point of view.
My observations are not meant to criticize what you said since
every computer system is also different and there will obviously
be a different set of priorities. All I wanted to do was to mention
how I handled the RAID 1 problems and my solution along
with reasons.
By the way, the PCI (non-RAID 1) controller is under $ US 100.
The 40 GBytes EIDE hard drives are under $ US 100 each.
Economics does, in my opinion, also play an essential role when
decisions are made.
If anyone else has any RAID 1 observations and suggestions,
I would like to read other points of view.
Sincerely yours,
Jerome Fine
--
If you attempted to send a reply and the original e-mail
address has been discontinued due a high volume of junk
e-mail, then the semi-permanent e-mail address can be
obtained by replacing the four characters preceding the
'at' with the four digits of the current year.
Received on Fri Apr 26 2002 - 11:39:00 BST