[Tfug] Hardware reliability

[Bexley Hall]

--- On Fri, 4/17/09, Judd Pickell <pickell at gmail.com> wrote:

> Although I can not speak the same for enterprise hardware I
> have worked with in my programming history. I have seen
> systems fail, although mostly just losing a hard drive or
> getting a bad memory stick. I don't think I have had a
> CPU completely fail on anything. Oh there was one that lost
> a redundant power supply, but that is hardly noteworthy.

CPUs rarely fail -- assuming they don't *melt*  :>
Usually, the power supply decoupling for the CPU is
the culprit if the CPU starts acting flakey.

I've never had a *hardware* disk failure.  Memory can
sometimes be traced to bad sockets (some of the sockets
for SIMMs/DIMMs are only rated at *6* insertions  :-/ )


--- On Fri, 4/17/09, Zack Williams <zdwzdw at gmail.com> wrote:

> I don't think I'd call you lucky - just that you've picked
> good quality components over the years.

But, even *that* is some amount of "luck".  E.g., I can
understand some "off brand" having quality issues.  But, I've
seen machines from IBM, Dell, HP, etc. with systemic problems!

> The failures come down to basically 3 things:
> 
> 1. Manufacturing defects (bad capacitors mainly)

I see a lot of cold solder joints.  This is especially problematic
in the SMPS regulators -- mainly the magnetics (though sometimes
recirculating diodes)

Note that in some consumer kit (i.e., not computer related),
there are actually *intentional* design defects.  I.e.,
those devices are *guaranteed* to fail quickly (like using
6.3V decoupling on 10V supplies)

> 2. Dead components with moving parts (fans, power supplies,
> hard disks, optical drives)

Fans have been my bane.  I've seen lasers fail on CD-RW's
but not usually the mechanism itself (at least not in the
units I have owned)
 
> 3. Physical abuse.  (mainly laptops that take a beating)

Ah, I hadn't thought of this!  :<  I rarely use my laptop
(unless I will be out of town) so it has very low wear.
I've seen laptops on which the legends of the keys have
worn off (this probably because they weren't two part
molds so the legends were "just paint")

> While you can generally prevent #3, #1 is really hard to
> identify other than retroactively as a group and #2 is pretty much
> random.
> 
> Some models are known to be worse than others - for example, 
> certain Dell towers and iMac G5's I automatically pop open at the
> first sign of trouble because they're know to have bad caps.

I had reservedly accepted the "industrial espionage" legendry
behind the caps problem.  But, it has continued far too long
to just be attributed to that.  Or, perhaps that was the
*cause* of the problem but manufacturers are just not
addressing it (i.e., I see more recent machines exhibiting
the same sorts of bad components).


--- On Fri, 4/17/09, Jim Secan <jim at nwra.com> wrote:

> I have never-ever-ever had any problems with the two Sun 
> systems I have.  One purchased in 1993 and still running, 
> the other in 1999 and still running (both running 24x7x365).
> Their hardware was great.  Maybe still is, but I've not bought any
> lately.

Note that 24/7 machines can give you a false sense of
confidence.  Many problems appear when a machine is
first powered up/booted (e.g., disks/fans that fail to
spin up)

In general, my Sun experience has been great.  With the
notable exception of a SunPCi2 (failing cooler and it
also looks like I will have to re-cap it -- something
I have never seen in a Sun product!)

I have had floppy drives in 4m machines "bind up" -- but,
that's almost forgivable given that you *never* use the
floppies on these things (they have a viscous lubricant
that gets gummy over time -- a simple fix but inconvenient
as it usually comes when you *need* to use the floppy...)

Amusingly, my biggest peeve is with the *feet* on many
Sun products!  The plasticizer seems to deterioriate in
the "rubber" and you end up with a gummy mess that rivals
ABC bubble gum!  (annoying if you sit that piece of kit
on any piece of furniture!)


--- On Sat, 4/18/09, John Gruenenfelder <johng at as.arizona.edu> wrote:

> At work it's only slightly different, and includes machines
> I've built and those built by others.  We came close to losing 
> an HD back in the P3 days (don't remember what year that was).
> I actually thought it was fan or PS fan noise, but it turns out
> that god awful sound was from the HD.  Oops.  :)  Got
> the data off it, though.

<grin>  I replaced, one by one, all of the fans in my U60
before finally realizing that wailing noise was coming from
the SunPCi2's cooler!  Grrrr... too hard to localize
those sounds in a confined space (especially when you
forget that there is a cooler in there!)


--- On Sun, 4/19/09, Malcolm Schmerl <mjs355 at comcast.net> wrote:

> For what it's worth, aside from the occasional noisy fan, I
> seem to go through quite a few power supplies. I'm wondering what
> could be the cause. I live out in Avra Valley. Our electricity is
> supplied by Trico.
> All my machines are on surge protectors. Could the input
> power still be the cause?

<frown>  "Surge protectors" are a great idea that doesn't
quite hold up to close scrutiny.

The components used in many of these aren't "big enough" to
handle any serious line disturbances.

When I lived in Chicagoland, one residence had notoriously
bad power -- an outage *every* month, etc.  Since coming
here, things seem much "cleaner" (and far fewer outages).
It woul dbe hard to make a sweeping generalization about
your particular situation as this can vary from one
neighborhood to the next (e.g., the folks on the next
block are on an entirely different "distribution circuit"
than we are; their power comes in overhead whereas ours is
below grade, etc.)

You'd have to do a post mortem on your power supplies to see
what the real cause is.  I've been seeing lots of power
supplies with bad electrolytics -- the same problem that
has plagued motherboards, etc.

Thankfully, these are usually easy/inexpensive to repair.
On the other hand, most power supplies are commodity parts
so you don't really $ave much in the process.