|
Interesting observation I have made over the last few years, specifically
on Dell servers is that the hotswap backplanes do not tolerate rapid and
frequent temperature changes. We have observed server rooms
that remain at a constant temperature over a few degrees, and even if the
temp is always 93 F, no backplane issues. However, in smaller rooms
with less thermal mass and with not enough thermostat hesteresis(too wide
of a trigger range), the temperature can vary by 10 degrees over 10
minutes. This can cause odd random backplane connectivity issues,
where drives seem to 'shake' off of the array.
I agree that the room should be much cooler, I
hate coming in on the weekends here, but the management has an "if
it ain't broke don't fix it" attitude and point out that we have had no
significant problems over 5 years so why change things now. We have had
a few drives (4 out of 20) fail over the years, some internal, some in a
Powervault, but nothing that seems out of the ordinary for 5 year old 10k
rpm drives that are always on. Since they are all raided, it has not
caused us any trouble yet and we simply replace the drive under our sevice
contract. I always look at it as an opportunity to get more drive space
as they don't make drives that small anymore.
Upgrading our drives one at a time.
:o)
4 failures out of 20 drives over 5 years.
Does that seem too high a failure rate or about average?
If it could be proven that the high temps are
causing drive failures the management might be a bit more interested in
upgrading the AC system in the computer room.
Doug
----- Original Message -----
Sent: Friday, August 12, 2005 11:30
AM
Subject: Re: [Declude.Virus] OT -
Server Room Temperature
Doug,
Hard drives are probably the most sensitive
components that you have in your servers, and I am not aware of any hard
drives that should be run above 50C/122F. My server runs about 35F
hotter for the system temp than the environment and about 40F hotter for the
CPU's than the environment. Note that these readings are under normal
load, but when the server redlines, the CPU's increase by about 15F and the
system by about 5F. Considering that the hard drives create heat
themselves and their much lower tolerance for heat in comparison to solid
state components, it would seem that going over 30C/85F for the ambient
temperature would be very dangerous as far as the hard drives go in an
active server. Hard drives will likely go over their operating
temperature long before the system or the processors unless you have a
broken fan or bad connection with a heat sync. My system is spec'd at
15C/27F over the hard drive's tolerance, and my CPU's at 27C/50F
over.
IMO, 66F is the proper server room temperature, and it gives
some leeway for adding more equipment and other issues that can crop up such
as A/C failures. 72F would be the high end normal temp that I would
want to see. If my colo was over 75F, I would definitely
complain. The guy next to me with 25 TB's of 15,000 RPM SCSI drives
would probably complain louder :)
Matt
Doug Traylor
wrote:
We just looked at the operating spec of our
servers from the Manufacturer's (Dell) website. The max is listed as
95* F and we run around 80* F during the day on weekdays and up to 92* F
on the weekends when they turn off the AC in the plant. We have our
own AC which runs 24/7 in the computer room/closet. So far we have
not had any noticeable system problems in the five years we have been
operating this way.
When we had a large IBM mainframe with all
the dressing, we kept it in a large computer room that was kept at a
chilly 66* F. I was a computer operator then and worked in there for
8-12 hours a day. I would wear two shirts and longs sleeves to
work, even when it was 110* F outside - Texas.
Doug
-----
Original Message -----
Sent:
Thursday, August 11, 2005 8:58 AM
Subject:
[Declude.Virus] OT - Server Room Temperature
Can someone point me to a source of
information regarding what temperature a server room should be at
?
Thank you.
|
