[Discuss] Tired of unreliable external drives... recommendations?

[Edward Ned Harvey]

> From: Bill Bogstad [mailto:bogstad at pobox.com]
> 
> On Sun, Jul 15, 2012 at 9:42 AM, Edward Ned Harvey <blu at nedharvey.com>
> wrote:
> >...
> > As an IT guy and Computer Engineer, and EE, working for CS/EE / chip
> > companies, I encourage companies to stick with big name brands where
> they've
> > sold millions of the same drive under warranty (such as ... buy your
Dell
> > and Apple and Oracle branded drives rather than OTS
> > seagate/westerndigital/etc drives.)  But it's all just a superstition
and a
> > balance of probabilities, and in fact, buying the supported name branded
> > drives does indeed cost more per drive.
> 
> None of the companies you recommend make their own disk drives.   In
> fact, they all resell drives from the companies that you don't
> recommend.   Now, I can remember a time when Sun (for example) had
> particular firmware revisions that they used for drives they resold.
> This would in fact make them different from the same model disk drive
> that I can purchase from NewEgg, etc.   Can you clarify what
> differences you believe still exist between branded and OTS drives
> when we are talking about the exact same model?  

For one, if you get a support contract on your server with
dell-oracle-whatever branded hard drives, then they become responsible for
managing the supply chain for replacement drives.  Three years later, when
model X disk drive isn't available from newegg anymore, you can still get
your replacement from dell etc, assuming you bought the dell etc up front.

For two, I formerly worked as a computer engineer for Compaq qualification.
Here's a summary of life in that job:  They hired me and a bunch of other
people into the qualification team for a product that didn't exist yet.  Our
job was to keep busy until the group 0 build was released to us...  And then
test the hell out of it as fast as possible.  It's guaranteed we'll discover
bugs, and management has a definition of "kit stopping" bugs.  It's pretty
simple, as long as the OS is able to boot and it's not a safety hazard, then
it's not a kit stopping bug.  (Both in the software and the hardware.)  (And
firmware.)

>From this experience, and many other experiences, I know and expect there
are bugs in every piece of hardware you own or operate.  In the best cases,
you might just power cycle your monitor to clear up a glitch, and in the
worst cases, your hard drive actually hasn't written data that it said it
wrote.  Or your CPU silently mis-computed something and you now have corrupt
data.

As a manufacturer, for example a hard drive manufacturer, you release your
product for sale, and for a period of time, you keep a team of engineers on
the job, fixing bugs.  After some time, you cease this effort.  Typically
around 6 months, but it varies by product, manufacturer, etc.  

I have in fact been hired at a company that built a compute farm of black
box linux machines, they hired me in part to figure out why the systems were
so unreliable.  I dug into it and found a bug in the on-board ATA
controller, that should presumably be fixed with a firmware update.  I
forget now if the motherboard manufacturer was Asus or another big name
(Asus comes to mind).  I contacted them, and they said they had reassigned
the staff formerly supporting this product to a different product.  They
would not be releasing an update for this problem.

The more millions of people have bought your specific product, and the more
they paid in aggregate for the support contracts, the more the manufacturers
care about actually fixing the post-release bugs.  Most of the post-release
bugs go undetected.  

When you buy a Dell (or whatever) rebranded drive, you're getting the
carefully selected commodity OTS drive that a team of engineers from Dell
vetted out and settled on.  You're getting post-release support that is
greater than the post-release support available to general public.

In theory, if you can figure out that Dell rebranded the Seagate XYZ, you
could go buy the Seagate XYZ directly and save money.  But this line of
logic bit me once:  I had an apple X-raid.  I got my hands on one of the
Apple-branded drives for the system, I figured out that the drives were
Hitachi XYZ, and I went out and bought Hitachi XYZ instead of apple branded.
When the hitachi drives arrived, they came in two batches with two different
FW revs.  Half of the drives worked flawlessly, and the other half could not
be recognized by the array.  Unfortunately, the ones that worked were the
ones with the older FW rev...  And the older FW rev wasn't available from
hitachi anymore.

I am not saying don't use commodity drives.  I know I use them.

I am saying, stick with enterprise products and enterprise contracts on
systems that are 24/7 critical.

Whenever I build a compute farm, and I can afford for 10% of the machines to
be down at any time, I buy the blackboxes and hire an intern to deal with
all the idiosyncracies.