ATA Raid

[Bill Gradwohl]

Although my initial request for info was to find out if any mobo based ATA
RAID chip sets were credible, that thread has now segued a bit but I'll
respond. 

BTW - It does look like I'll get a sample 3WARE and RAIDCORE to test with
since it doesn't look good for the mobo RAID chips given the responses.

> Kent wrote - 	
	Don't blame Linux for the sins of other OSs.

I'm not blaming anything - I'm stating the obvious. All OS's have bugs. The
software flashed into a RAID chip can also have bugs, but in the 20+ years
of setting up SCSI RAID arrays, I've never had a RAID controller fail either
via hardware or its eprom software. Backplanes have failed, which is why I
avoid them, but the controllers and their on board chips have never given me
cause for concern. 

The O/S is simply a richer source of failure. RAID logic has been tested
over lots of years and is a narrowly focused collection of code. When its
burned into a chip, its largely immune from corruption by viruses, rogue
ring 0 code, and even hackers.

ATA RAID logic has to be very similar to SCSI RAID logic at least at some
level. Therefore ATA RAID inherits credibility from the SCSI work done that
preceded it as far as I'm concerned.



Your description of how RAID works also applies to the hardware RAID variety
so its a wash.

> Kent wrote - 	
	I once had a disk die ... and I never noticed ...until the next
morning ...

That's a problem. Many hardware RAID controllers scream when something goes
wrong. The machine beeps, whistles or in some way is annoying to attract
attention. It allerts the sysadmin to a problem, the solution to which is
varied. It can be to do nothing if he has a hot spare (which doesn't require
a backplane) that's been in the machine since array creation, or it can be
to phone in an order for a replacement drive. It depends on the situation. 

> Kent wrote - 	
	Linux software raid 1 is even bootable--though getting your BIOS to
try more booting from than one disk might be tricky

Hardware RAID is always bootable (given you got the right controller in the
first place) even if I remove a physical drive in a RAID 1 or 5 array. No
worries about the machines BIOS not wanting to use an alternate boot volume
because it always uses the same boot volume - the array as the array
controller wants to present it, even in a degraded state.

> Kent wrote - 	
	One key point of raid 1 is reliability, and reliability likes
simplicity and eliminating single points of failure

I agree in principle. However, I also like redundant power supplies. They
are far more complex than a non redundant version, but I've also never seen
one fail as a whole in 20+ years. A portion of a redundant supply may fail
but the O/S on the machine never noticed. So, reliability and simplicity do
not go hand in hand 100% of the time. 

> Kent wrote - 	
	and getting things back up and running quick when they do fail.

With a hot spare I am forced do absolutely nothing when a drive fails. If I
want, I can replace the failed drive whenever I get around to it. If the box
is already old enough and due to be replace in 12 to 18 months, I wouldn't
bother. I'd let it run with no hot spare, but with complete RAID redundancy.
One more drive can fail and I'm still in good shape. 

> Kent wrote - 	
	Are you going to quit running any software?

No - I have no intention of scrapping my servers in favor of a Big Chief
Tablet and a number 2 pencil. 

> Kent wrote - 	
	The software in Linux raid code is very well tested by now--will
that be true of the embedded software in your "hardware" controller?  

I believe the RAID knowledge base Adaptec and its offspring RAIDCORE and to
some extent 3WARE are using predated Linux by at least a decade. I trust
them more.

> Kent wrote - 	
Will needed firmware updates be as promptly released?

To date, I've never cared if they released firmare updates to the RAID
controllers I've used. In 20+ years, I've never flashed an operational
servers RAID proms. I've never had to - the code has been that good. I
usually flash the mobo BIOS and the RAID BIOS to the most recent versions
before I installand then thoroughly test the O/S & RAID combination. That's
before it goes into production. I have patched vendors RAID drivers for a
particular O/S on a production server, but was forced into that only once.

-- 
Bill Gradwohl
YCC
(817) 224-9400 x211
www.ycc.com 
SPAMstomper Protected E-mail
www.stomperware.com