[Discuss] RAID question

[Patrick]

On Mon, 2 May 2016 21:17:26 -0700
Drew wrote:

> > That was
> > my other main concern about software RAID; it seemed to dump a
> > hardware job on the software, with all the usual performance
> > penalties.  When software tries to emulate hardware, it always
> > slows things down by some fraction.
> 
> I'm going to call this statement out as erroneous. Modern CPU's are
> overpowered for a lot of work loads and the spare CPU horsepower gives
> hardware RAID a run for it's money.

Okay, good point.  I'd been using low-end computers for a while,
and saw little reason to change that because as you said, modern
CPUs are overpowered for most tasks.  So the old stuff does still
work for most things, although lately I've been noticing its
limitations more and more.

For that reason, maybe I'm too easily impressed -- although
once again I did buy and old computer.  Just surprised how
*much* of a computer I could get for $200 .

> Utilizing the CPU also permits
> larger cache sizes (most HW RAID has 1-2GB cache max, mdadm (linux
> software raid) can access a good portion of main RAM) which improve
> performance by an extra margin.

Well yeah, faster access does speed things up a lot [something
I'd noticed after sticking an SSD in my laptop], and access to
memory is faster than *any* kind of disk available these days.

[...]
> 
> I will also note there is a distinction between software RAID, and
> fake-RAID (software assisted RAID). Software RAID, like mdadm, is pure
> software. As long as the linux OS can see the drives, mdadm can RAID
> them. Fake-RAID on the other hand hides the individual drives behind a
> raid controller but offloads the actual work to the CPU. Fake-RAID is
> what you usually see on consumer PC's with 'RAID' supplied by a bios
> switchable drive controller. And Fake-RAID is in many ways the worst
> of each. :)

...Won't tell you what it's doing, but uses system resources to
do it.  Got it.

On a similar note, my MP3 player developed a minor problem where
its power-off timer set itself to 0 -- meaning that if you paused
it, it would shut down as soon as the backlight went off.  Kind
of annoying, but I got in the habit of unplugging the headphones
instead of stopping playback, and the not knowing what would be
playing when I plugged them back in was kind of fun.  :-)

Then just last week, it developed another problem where it
completely forgot how to turn itself off.  This corrected the
first problem, but in both cases I have no idea why [except to
guess that maybe that one byte of flash memory where it used to
store its power-off setting had failed somehow, and couldn't be
rewritten].

Anyway, that two-slot RAID box arrived, and it does work, but
yeah, I'm not sure *how* it works.  Not too worried about it
failing though, or either drive failing, because in RAID 1 mode
it produces two identical, ordinary SATA drives which can be
removed and used separately [I've been testing it].

It's compact, easy to carry, and can be OS-agnostic depending
how you format its partition[s].  Found that it draws very
little power as well [also tried running both the power and
data lines off a single USB bus], and should draw even less
when the SD/RAID 0 adapters get here.

Even *more* compact would be a USB hub loaded with the smallest
flash drives available [physically smallest, not talking about
storage capacity].  Probably much more reliable too, but not at
all OS-agnostic.

> "This started out as a hobby and spun horribly out of control."
> -Unknown

Hey, I resemble that remark!


Cheers,
Patrick.
-- 
We don't understand the software, and sometimes we don't
understand the hardware, but we can *see* the blinking
lights!