2

I have two temperature readings for my server room one measures the ambient temperature the other measures the exhaust for a server cluster. I have noticed recently that the difference between these temperatures can reach near 30F is this to large?

Our server room houses a large server cluster, about a dozen rack mounted servers and is cooled by this a/c unit, The estimated maximum total output from the servers and lighting is 121110btu. There is no air exhaust in the room, I was thinking that adding an exhaust would help but was looking for a second opinion.

Images of temperature readings:

Low usage Low Usage

High Usage High Usage

4
  • 2
    Looks like you accidently uploaded the same image two times. Feb 3, 2020 at 14:19
  • Hi, I like sensors reporting, but is it actually cold or hot in the room if you walk by there ? You miss that information.
    – yagmoth555
    Feb 3, 2020 at 14:26
  • @yagmoth555, The room does not feel hot but the exhaust from the cluster is quite warm
    – David
    Feb 3, 2020 at 14:34
  • @GeraldSchneider Thank you, fixed.
    – David
    Feb 3, 2020 at 14:36

1 Answer 1

1

This would seem "within the bounds of normal" - maybe slightly on the warm side depending on how your ventilation is set up and where its measured. It likely also depends on the servers - more power hungry servers = more heat to dissipate.

https://www.stulz.de/en/newsroom/blog/delta-t-91/ is an interesting links talking about temp differences in the SC, and reflects 10-15 degrees C as normal.

Alternatively think of temp differences in hot/cold isle DC or what its like to walk in front if a heater. Remember that a cluster if servers drawing 1kw continuously is more-or-less equivalent to a 1kw heater - and what does it feel like when you walk past one of them.

Lastly, temperature (while important) is probably less important then changes in temperature from a reliability POV

6
  • Thank you for your answer, by changes in temperature would a safe range also be 10-15 degrees C?
    – David
    Feb 5, 2020 at 13:39
  • Yes, but I think you misunderstand what I'm sating. Its not the difference between input and output temperature, its the difference in input temp over time - so (for example) having the room at 15c at night going to 40c during the day will put more stress on components then having it consistently at say 23c. (ie an air exhaust might be a step in the wrong direction - especially as it implies an air intake which means less control of incoming temp, humidity and probably more dust, Depending on your goals YMMV though.)
    – davidgo
    Feb 5, 2020 at 19:06
  • Yes I understand that you mean the difference over time in the server room, our issue currently is that hot hair seems to pool behind the server and there is nowhere for it to go. But based on your answer it seems that our current max temp of 95F(35C) is not something that we should be worried about?
    – David
    Feb 6, 2020 at 17:07
  • @Bjorn Don't understand? Where are you measuring the output? If you were (mentally) replacing your servers with a blow heater - which is what the servers are acting as - wouldn't you expect the output to feel warm? This all sounds normal to me.
    – davidgo
    Feb 6, 2020 at 19:17
  • We are measuring the temperature right behind the server, of course we expect it to be warmer than the air going in, I was questioning if the temperature we were reading was too high. As you mentioned it does not matter so much what the temperature is so much as the change in temperature over time is. I worry that our change over time is too high. As you can see in the graphs above we are experiencing a 20 degree(F) change over about a 3 hour time interval, and this happens nearly every day. Sorry if I was not clear in my original questions.
    – David
    Feb 7, 2020 at 20:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.