On 25 Feb 2004 18:21:21 -0800, (Zymurgy)
wrote:

If you don't take any steps to control the temperature then it could be.

Methinks you worry too much.

Any decent kit (i.e something that's not been cobbled up in a 3rd
party case) has a wide operating temperature, and it is the rate of
change of the ambient temperature that causes an issue, not the
actual maximum or minimum temperatures.

I regret that I disagree with you - using an analagy it's like you
saying that your car is fitted with seatbelts so you are immune from
being involved in an accident!

Equipment safety functions can - and do - go wrong. So whilst your PC
(etc) might work just fine in a given temperature range the scope
might change very markedly if say the internal fan fails. By way of
example if the PC is operating within its safety limits with an
ambient temperature of 35 degrees C and then an internal component
goes out of spec then a temperature in the equipment could rise
markedly and lead to a fire.

I simply don't think that having a single point of failure within
equipment is a good idea, if you are continuously relying upon it to
save your bacon.

Your home insurance company might well take a dim view of paying out
if your equipment were to catch fire. Whether or not they can be
challenged on their decision is another matter - but you can feel
comfortable that they can afford a bigger team of lawyers than you
can.

PoP

-----

My published email address probably won't work. If
you need to contact me please submit your comments
via the web form at http://www.anyoldtripe.co.uk

I apologise for the additional effort, however the
level of unsolicited email I receive makes it
impossible to advertise my real email address!

(Zymurgy) wrote in message . com...
Andy Hall wrote
PoP wrote:
Jason Arthurs wrote:

My server currently runs in the loft
I am curious!

I would have thought that running any sort of equipment in the loft is
asking for trouble. Reason being that the temperature up there can go
outside the specs of the equipment. It gets damned cold up there in
winter and sweating hot in the summer.

If you don't take any steps to control the temperature then it could be.

Methinks you worry too much.

Any decent kit (i.e something that's not been cobbled up in a 3rd
party case) has a wide operating temperature, and it is the rate of
change of the ambient temperature that causes an issue,
not the actual maximum or minimum temperatures.

That's rubbish. Electronic equipment can and does fail when operated
outside of its specified temperature range. The spec for any piece of
equipment is a worst case figure and you may well be able to operate a
particular example well outside that spec but there will be a limit at
both high and low temperature.

My firewall (Compaq Pentium 2) is in the loft with no special
insulation around it, and the projected low for tonight is -4.

The low temperatures we get in this country (even left outside) would
not generally be a problem for most electronic equipment if left
powered up to keep it warm.

I have had computer room air conditioning failures causing a very
steep step change in temperature (up to ~32 degrees C in 1 case) I had
3 disk failures out of an installed base of ~600 and 1 machine auto
shut down due to overtemperature out of 200 (and this was found to be
too close to adjacent kit).

So? The air con failed and the ambient temperature became too high. It
has nothing to do with rate of change. Even if you had raised the
temperature over the space of a week you would have seen the same
failures.

Similarly I have seen machines operate at either end of their
envelopes [1] without incident.

That's what they're designed to do. Take them outside their envelope
and they *will* fail eventually.

Andrew

Andy Hall wrote in message . ..

Methinks you worry too much.

Any decent kit (i.e something that's not been cobbled up in a 3rd
party case) has a wide operating temperature, and it is the rate of
change of the ambient temperature that causes an issue, not the
actual maximum or minimum temperatures.

My firewall (Compaq Pentium 2) is in the loft with no special
insulation around it, and the projected low for tonight is -4.

I'm not worried about it. It will just get extra cooling this evening
;-)

I have had computer room air conditioning failures causing a very
steep step change in temperature (up to ~32 degrees C in 1 case) I had
3 disk failures out of an installed base of ~600 and 1 machine auto
shut down due to overtemperature out of 200 (and this was found to be
too close to adjacent kit).

Similarly I have seen machines operate at either end of their
envelopes [1] without incident.

Cheers,

Paul.

[1] Spec sheets on mine say:
5 to 35 C Operating
-20 to +60 Non Operating

I don't disagree that rate of change can have a big effect, although
high continuous temperatures do affect the failure rate of electronics
in general. For example, it is common to do accelerated tests at high
temperatures to check for failure in new designs.

This kind of testing is looking for permanent, physical, failure
modes.

Operating at too high a temperature in a loft is unlikely to do
permanent damage (unless the temp is really extreme) but will cause
malfunction.

Andrew

PoP wrote
Zymurgy wrote:
Methinks you worry too much.

Any decent kit (i.e something that's not been cobbled up in a 3rd
party case) has a wide operating temperature, and it is the rate of
change of the ambient temperature that causes an issue, not the
actual maximum or minimum temperatures.

I regret that I disagree with you -

Well, it'd be a boring group if we all agreed on something ;-)

using an analagy it's like you saying that your car is fitted with seatbelts so you are immune from
being involved in an accident!

Not saying that at all, just that using my not insignificant knowledge
of computer management in large datacentres,
across an enormous range of kit i'm just saying that measures such as
these from Andy:

:I dealt with it by making an insulated cabinet and arranging two fans
:with ducting in and out from the outside and inside the house. The
:fan speeds are controlled by a temperature sensor and motor
:controller, and there are servo controlled dampers in the ducting.

are overkill in my mind, and still do not mitigate for equipment
internal cooling fan failures.

Your home insurance company might well take a dim view of paying out
if your equipment were to catch fire. Whether or not they can be
challenged on their decision is another matter

Equipment will catch fire if it is so predisposed wherever it is
located.

If this was caused by loft overtemperature, then I agree, there will
be some culpability. If we have a long hot summer reminiscent of '76
then I will consider secondary cooling, or relocation of the
equipment, until such time, I remain nonplussed.

I remain convinced my firewall will be up after its freeze thaw
session last night, due in no small part to the microclimate around it
;-)

Cheers,

Paul

On 27 Feb 2004 04:03:11 -0800, (Andrew)
wrote:

Andy Hall wrote in message . ..



I don't disagree that rate of change can have a big effect, although
high continuous temperatures do affect the failure rate of electronics
in general. For example, it is common to do accelerated tests at high
temperatures to check for failure in new designs.

This kind of testing is looking for permanent, physical, failure
modes.

Operating at too high a temperature in a loft is unlikely to do
permanent damage (unless the temp is really extreme) but will cause
malfunction.

I know.

I was using this by way of example rather than as a cause. However,
higher temperatures, even on average do increase failure rates.

The main point of maintaining suitably low temperature, as you say, is
to prevent malfunction.


..andy

To email, substitute .nospam with .gl

On 27 Feb 2004 10:22:14 -0800, (Zymurgy)
wrote:



Not saying that at all, just that using my not insignificant knowledge
of computer management in large datacentres,
across an enormous range of kit i'm just saying that measures such as
these from Andy:

:I dealt with it by making an insulated cabinet and arranging two fans
:with ducting in and out from the outside and inside the house. The
:fan speeds are controlled by a temperature sensor and motor
:controller, and there are servo controlled dampers in the ducting.

are overkill in my mind, and still do not mitigate for equipment
internal cooling fan failures.


Well... yes and no.

Datacentres use air conditioning, at not inconsiderable cost, to
maintain a realtively constant low temperature because the equipment
manufacturers suggest that reliability in the short and long run is
improved.

A branded product will typically have higher quality components and
more attention will have been paid to environmental factors than might
be the case in a Hu Flung Dung superspecial.

The internal temperatures (which is ultimately what counts) of PC and
more sophisticated networking equipment can be monitored and logged
and used to trigger shutdowns in the event of problems. This might
be through ambient temperature rising or specific equipment fan
failure. I tend to fit extra cooling fans to equipment and power it
separately to the equipment PSU anyway. This is a pretty cost
effective way of working because fans are cheap.

Temperature monitoring inside is easy enough to do as well, and it is
also simple to look for sudden rates of change of temperature and
excessively high temperatures and to shut things down. For my
particular requirements, I can selectively do that and still have
enough redundancy to keep working as I need to do.

Using controlled external cooling air is quite common for equipment
racks as a means of maintaining a reasonablly constant temperature and
keep things working in circumstances where the high temperatures would
otherwise result in malfunction or failure. As always, it's a
cost/benefit trade off, but I've found that the way I've done this
works pretty well in terms of implementation and running costs.






..andy

To email, substitute .nospam with .gl

On Fri, 27 Feb 2004 21:11:30 +0000, Andy Hall
wrote:

Datacentres use air conditioning, at not inconsiderable cost, to
maintain a realtively constant low temperature because the equipment
manufacturers suggest that reliability in the short and long run is
improved.

The air conditioning in datacentres usually also performs
environmental scrubbing operations - removing dust particles from the
air.

If dust were allowed to continue circulating then the equipment
motherboards (etc) would over time acquire an overcoat of dust. That
could potentially cause localised overheating at chip level.

However I digress. I'm showing my own long-term knowledge of working
in a maintenance role in datacentres

PoP

-----

My published email address probably won't work. If
you need to contact me please submit your comments
via the web form at http://www.anyoldtripe.co.uk

I apologise for the additional effort, however the
level of unsolicited email I receive makes it
impossible to advertise my real email address!

Andrew wrote:

(Zymurgy) wrote in message . com...

Andy Hall wrote

PoP wrote:

Jason Arthurs wrote:


My server currently runs in the loft

I am curious!

I would have thought that running any sort of equipment in the loft is
asking for trouble. Reason being that the temperature up there can go
outside the specs of the equipment. It gets damned cold up there in
winter and sweating hot in the summer.

If you don't take any steps to control the temperature then it could be.

Methinks you worry too much.

Any decent kit (i.e something that's not been cobbled up in a 3rd
party case) has a wide operating temperature, and it is the rate of
change of the ambient temperature that causes an issue,
not the actual maximum or minimum temperatures.


That's rubbish. Electronic equipment can and does fail when operated
outside of its specified temperature range. The spec for any piece of
equipment is a worst case figure and you may well be able to operate a
particular example well outside that spec but there will be a limit at
both high and low temperature.



I have to agree. Ther are failure modes associated with temperature
cycling, most;y mechanical stress leading to failures of joiints and
frit seals on chips, but by far and away the usual cause of
semiconductor PERMANENT as well as TEMPORARY degratation is overtemperature.



My firewall (Compaq Pentium 2) is in the loft with no special
insulation around it, and the projected low for tonight is -4.


The low temperatures we get in this country (even left outside) would
not generally be a problem for most electronic equipment if left
powered up to keep it warm.



Most commercial equipment can reasonably be expected to work between 0C
and 40C. The chips themselves are generally in spec between -5 and 70C,
but that is not the whole story....MIL spec stuff is rated between -25C
and 125C



I have had computer room air conditioning failures causing a very
steep step change in temperature (up to ~32 degrees C in 1 case) I had
3 disk failures out of an installed base of ~600 and 1 machine auto
shut down due to overtemperature out of 200 (and this was found to be
too close to adjacent kit).


So? The air con failed and the ambient temperature became too high. It
has nothing to do with rate of change. Even if you had raised the
temperature over the space of a week you would have seen the same
failures.



Precisely. Internal air temps over 50C are almost certainly indicative o
very high junction temperatures - go ovcer 175 junction on MOS and its
'good night, vienna'.



Similarly I have seen machines operate at either end of their
envelopes [1] without incident.


That's what they're designed to do. Take them outside their envelope
and they *will* fail eventually.



Mostly thety stop working before they fail. Chips are made to lie
withing specs, but no manufacturer in the world designs his kit to
accept components that are all at the worst possible end of the
specification spectrum.

Insdtead a monte carlo analyisis is done at best. In practice what
actually happens is that the designers do their best, a few prototypes
are temperature tested, and the production goes ahead. If lots of users
report a similar problem then the design may be examined, but mostly
they just get replacement boards. Its cheaper.

Even MIL spec kit os not necessarily designed to any different
standards, but it may well be sample tested in an envoironmental chamber
to ensure it works over the specifed range.


Andrew

Zymurgy wrote:

Equipment will catch fire if it is so predisposed wherever it is
located.

If this was caused by loft overtemperature, then I agree, there will
be some culpability. If we have a long hot summer reminiscent of '76
then I will consider secondary cooling, or relocation of the
equipment, until such time, I remain nonplussed.


The coolest place without forced cooling on a hot summers day is in open
shade. IF you have efficient loft ventilation, that is often the loft.

If there is restricted airflow and e.g. a tled roof, then the best
option is to buld an insualted room in teh loft, and arrange significant
ducts to it and force air through. Air temp seldom exceeds 30C in this
country above street level, and altho this is on the high side for
consumer equipment, its not a huge problem if there is an adequate
supply of it.



I remain convinced my firewall will be up after its freeze thaw
session last night, due in no small part to the microclimate around it
;-)



Low temperatures are not such a problem. Most kit will do -5C all right.
Semiconductors lose gain and get slower as they get colder. Sometmes
this is enoug to cause timing errors, and a system crash. Mostly its
higher temps that do the harm tho. I have fixed several recalcritant
servers by blowng ****e out of/replacing the fans and getting them
working again. Some died permently. Thse were all SUN SPARCS BTW.


Cheers,

Paul