Jump to content

Recommended Posts

Posted (edited)

Hi all, I'm looking for your opinions on my server room temperatures.

 

There is no air conditioning in the room. There is a vent grille on the door and an extractor fan in the ceiling.

 

I have 3 physical tower servers, a NAS and a comms cabinet with two switches a Ruckus ZoneDirector and a PoE midspan.

 

My cheap sensor from Aldi says 30c in the room and 33c inside the comms cabinet.

 

My server temperatures are:

Server 1 - system board 31c, CPU1 61c, CPU2 64c

Server 2 - system board 29c, CPU1 50c

Server 3 - system board 29c, CPU1 56c, CPU2 59c

 

In the comms cabinet, the temperature on my Procurve 2920 is 41c, 32c on the NAS.

 

The maximum limit for my switches is 45c. 40c for the ZoneDirector and 40c for the midspan PoE so I'd say they are very close.

 

It's only 12c outside here at the moment, if we get a sustained period of 20c+ the school gets unbearably hot.

Edited by JordanT91
Posted

Thats far too hot IMO.

 

We run ours at 21°C, and got an alert if it goes above 23°C.

 

Youll struggle in summer, its not even warm yet and you're up to 30°C in the room.

  • Thanks 1
Posted
There's a sweet spot - 14°C is going to be expensive to run and is not really necessary for longevity. Ours is set at 19 or 20°C, and the equipment is lasting well though it is a large room with good circulation. There has been some recent research on optimum temperatures in data centres, many have found that they can run a fair bit hotter with no/minimal reliability impact. I'd expect 25°C to have no issues really. Humidity is also important.
  • Thanks 1
Posted
Hi all, I'm looking for your opinions on my server room temperatures.

It is mostly about risk. How do temperatures affect reliability and what are the consequences if temperatures are a factor in equipment failure.

 

So Page 27 suggests that your chance of a server failure at 30 deg C is 1.42% compared to my servers running @ 20deg C which are at 1%. So actual chance of failure should still be relatively low but does increase as a function of temperature. Risk calculation is usually a product of probability *and* the consequences - what actually happens if something does break. Usually with our kind of infrastructure, failure is a matter of cost (rather than say lives) - how much does it cost if we are without a server for the time it takes to replace it? IMO this is quite difficult in a school. Costs tend to be low or are difficult to quantify (in a business getting a cost is relatively straightforward). We are still relatively good at dealing with/absorbing disruption but prolonged disruption will be unwelcome, with prolonged anything longer than a day or two. So it might be that elevated temps are not a hugely substantially elevated risk (compared to lower temps) unless you have very poor DR plans for hardware failure or perhaps you are invalidating the next day warranty cover because of the environment (which would be a poor DR plan for server failure!).

  • Thanks 1
Posted
You could reduce overall temps by figuring out the flow of air in the room. ie if you point the front of each device toward the door vent grille so air flow comes in and then is sucked up by the extractor fan you could see a noticeable drop in temps. not ideal if it's sucking in warm air but you could always use a fan?
  • Thanks 3
Posted
You could reduce overall temps by figuring out the flow of air in the room. ie if you point the front of each device toward the door vent grille so air flow comes in and then is sucked up by the extractor fan you could see a noticeable drop in temps. not ideal if it's sucking in warm air but you could always use a fan?

Unfortunately the way the room is laid out the servers are currently in the best possible place. There's so little airflow that you can't feel it.

 

I'll see about getting some AC in the room.

 

Thanks for all the advice!

Posted

A/C in server rooms, comfortable room temperature is 20 for humans, computers just want *consistency* not any one particular temperature.

 

Temperature changes are the enemy, really.

 

My rooms have got to the 30's when A/C has failed before (nobody ever has UPS/generators for that!), at which point I expect stuff to start warning and/or shutting down.

 

I don't care about the temperature inside the cabinet so much - I don't directly measure it. The room temperature and the power of those devices determine that.

 

Exhaust temps on blade-servers in A/C rooms are currently 20-25. It's 22 in my (non-AC) office.

 

The thing I hate is the people who come into the A/C rooms in the summer and go "Oh, isn't it lovely and cold in here" and then shelter in there, but in the winter are found saying "Oh, it's so warm in here, how do you work in here". When the temperature varies by less than a degree all year round. Just proves to me that all the "my classroom is cold" stuff should just be ignored.

  • Thanks 1
Posted

our server room AC is set to about 18C.

To be fair I've popped over to do 'server maintenance' in the summer when our south-facing office hits about 30 degrees.

  • Thanks 1
Posted
The problem will be convincing the Head to purchase the aircon. We've had one disk failure (a month ago) and no failures of any other kind in the nearly 3 years I've worked here. There's 1 extra tower server and 2 extra devices in the comms cabinet so the room is a little hotter but it was roughly the same temperature in 2014. Historically aircon has been a bit of a taboo subject in the school!
Posted
The problem will be convincing the Head to purchase the aircon. We've had one disk failure (a month ago) and no failures of any other kind in the nearly 3 years I've worked here. There's 1 extra tower server and 2 extra devices in the comms cabinet so the room is a little hotter but it was roughly the same temperature in 2014. Historically aircon has been a bit of a taboo subject in the school!

 

Just put it in writing that it is industry best practice to keep the ambient air temp under 25 degrees and that you recommend the immediate purchase of air conditioning for the server room in order to reduce the risk of premature hardware failure.

 

If they then ignore it and something bad happens then the failure is on them as senior leaders. It's classic poor leadership to assume the server room doesn't matter.

 

Have you got a local education authority you can get official advice from as they'll probably say the same thing, fit air con.

  • 3 years later...

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...