Lately it's been about 100 and humid in socal and the server room is way too hot. I'm wondering if anybody has any ideas to improve flow. It's about 5 racks with a few routers, a bunch of 2560g switches, and a bunch on supermicro 1u servers.
server room AC is maxed out but it's still getting way to hot (socal)
- Thread starter holden j caufield
- Start date
Get a portable unit. They make them for data center environments. But it sounds like it's time to upgrade the AC in the room if it can't keep up.
Is it a raised floor? Seal the floor as much as possible. They make brushes and things you can put over the holes where the cables go through to keep the air from escaping.
Is your cabling in the back of the racks clean? If not organize them and it will improve airflow.
Also, start using VMware!
Also, start using VMware!
They are elluding to the fact that you can consolidate all of the equipment in those 5 racks to much less equipment if you are running Virtual Servers. For example, a rack full of 42 1U Servers could potentially be reduced down to 5 - 7 Servers (obviously this would depend on the applications/server roles, and hardware resource requirement for each server). Less Servers/Hardware = Less Cooling Requirement.
In practice the opposite is true. Blades and VMware put out TREMENDOUS LOCALIZED heat. You can always tell what racks are blades/vwmware just by standing behind them, they blast out super hot air.
It is much more difficult to handle the cooling needs of blades/consolidation from a data center design perspective.
I don't work in a large data center (or even a medium sized one), so I can't relate to this. But, the OP mentioned having 5 racks, which is a pretty small footprint. I'm not sure that the localized heat would apply in this case, because all of the racks are probably within 15 or 20 feet of eachother anyways.
We went from ~65ish servers across 3 racks to ~15 servers in 2 semi empty racks.
(and that number is likely to drop below 10)
(and that number is likely to drop below 10)
The obvious: more AC or less gear. Thermodynamics is a harsh mistress.
The expensive: virtualize (check thermal ratings), rent offsite servers during the hot months
VM servers don't have to be hotter, depends on how much oomph you need.
The labor intensive: manage airflow (clear obstructions, stack blade servers together to prevent turbulence, ensure vented floor tiles are in front of racks), cover your black tar roof with a white tarp, etc
The weird: pipe in air from a cave, have 400lbs of ice delivered every day and blow fans across it, mount thousands of surplus peltier coolers in the walls...
The expensive: virtualize (check thermal ratings), rent offsite servers during the hot months
VM servers don't have to be hotter, depends on how much oomph you need.
The labor intensive: manage airflow (clear obstructions, stack blade servers together to prevent turbulence, ensure vented floor tiles are in front of racks), cover your black tar roof with a white tarp, etc
The weird: pipe in air from a cave, have 400lbs of ice delivered every day and blow fans across it, mount thousands of surplus peltier coolers in the walls...
Brovane
Diamond Member
- Dec 18, 2001
- 5,850
- 2,148
- 136
Are you tracking the temperature in your equipment? If you can start pulling some SNMP temperature MIB's and start tracking it. I have always found in these situations it is better to deal with cold hard facts than the Server room is hot. Figure out your temperatures and compare to what the optimal operating spec's.
If you wanted to get real exact match up the BTU's of cooling vs the BTU's of heat you put out.
Also if your AC unit(s) are not keeping up when all are operating then you will have a really big issue when something fails.
Oh it sure does. If the air is too dry, you run the chance of greater ESD the more common of the two. If the air is too moist, it doesn't cool as well and you can also run the risk of condensation, particularly if the ac is set to cycle off at night or anything that will change the temperature quickly.
Oh, and second for teh portable airconditioner. I have a small IT room that was a janitor's storage closet. It has no ducting for normal AC. I have a portable ac with the exhaust and cut a hole in the wall to put the heat exhaust and the water drain line into the next room which is the actual janitor's closest with the floor sink.
Of course, his room is hot as all hell, but he shouldn't be hanging out in there anyways...
Brovane
Diamond Member
- Dec 18, 2001
- 5,850
- 2,148
- 136
Yes it does. That is why at a good center the humidity is monitored and controlled.
http://searchdatacenter.techta...80_gci1261265,00.html#
Jeff7181
Lifer
- Aug 21, 2002
- 18,368
- 11
- 81
A computer doesn't get hotter because it's humid. People only feel hotter because high humidity reduces the ease at which your sweat evaporates which is a cooling mechanism of your body.
Of course you wouldn't want fog in your data center... but the temperature of your equipment is not going to change because the humidity level in your data center went from 40% to 60%. Brovane's link states officially, 20-80% relative humidity is fine. Anything outside this range is unacceptable not because it would make the equipment too hot, but rather because you don't want wet components or excess static. I think you'd be hard pressed to find a temperature controlled room where the humidity falls outside that range regardless of what the level is outside. Dehumidifying is a byproduct of cooling... it's possible for it to get too low... but the implication in the original post is that the high temperature and high humidity is making his equipment unhappy.
OP... your problem is purely heat... not humidity. (I welcome anyone that can prove me wrong)
And by the way...
:thumbsup: This is the best answer. Unless you're interested in extravagant methods such as freezing water overnight when electricity is cheaper and using it to help cool the data center during the day when it's hot and electricity is more expensive.
Just keep in mind... an AC unit displaces heat. It does not eliminate it. You need somewhere to dump the excess heat (outside).
Of course you wouldn't want fog in your data center... but the temperature of your equipment is not going to change because the humidity level in your data center went from 40% to 60%. Brovane's link states officially, 20-80% relative humidity is fine. Anything outside this range is unacceptable not because it would make the equipment too hot, but rather because you don't want wet components or excess static. I think you'd be hard pressed to find a temperature controlled room where the humidity falls outside that range regardless of what the level is outside. Dehumidifying is a byproduct of cooling... it's possible for it to get too low... but the implication in the original post is that the high temperature and high humidity is making his equipment unhappy.
OP... your problem is purely heat... not humidity. (I welcome anyone that can prove me wrong)
And by the way...
:thumbsup: This is the best answer. Unless you're interested in extravagant methods such as freezing water overnight when electricity is cheaper and using it to help cool the data center during the day when it's hot and electricity is more expensive.
Just keep in mind... an AC unit displaces heat. It does not eliminate it. You need somewhere to dump the excess heat (outside).
vailr
Diamond Member
- Oct 9, 1999
- 5,365
- 54
- 91
If you have access to an underground water supply: water cool the CPU's using a closed thermal loop.
Also known as: geothermal cooling.
Also: the relative humidity DOES affect the rate of heat transfer done by air cooling. I imagine humidity somewhere between 10% and 25% would be ideal.
Also known as: geothermal cooling.
Also: the relative humidity DOES affect the rate of heat transfer done by air cooling. I imagine humidity somewhere between 10% and 25% would be ideal.
Let me state that I am not in IT or have any knowlege about servers, but I do know about aerodynamics. I can see humidity affecting cooling a couple ways. First off let me compare a coolng fan in a server to the prop of an airplane. Airplanes are rated for speed and climb at sea level at a certain temperature and humidity. When you increase any of those the prop and engine lose efficiency. High humidity expands the air and reduces the propellers bite in the air, reducing flow. Secondly since these units are designed to transfer heat to air inside the server chassis, it is possible that the higher thermal capabilities of water reduce the transfer of heat to those molecules.
Or maybe I'm just blowing smoke.
Or maybe I'm just blowing smoke.
Cox colocation facility fixed this for us, it seems like they have monitor also that once it hits a certain temp an additional AC kicks in so luckily I didn't have to do anything
Jeff7181
Lifer
- Aug 21, 2002
- 18,368
- 11
- 81
While I don't doubt it has SOME kind of effect, I've yet to see anyone mention anything about how much of an effect it has. Considering everything I've read and heard indicates as long as there's no condensation it's not too humid, I doubt whether the effect is has is significant enough to even consider in this type of environment. I mean really... are we talking a 20% decrease in cooling at 60% humidity vs. 40%? Or is it more like 2%?
When you're talking about air planes, especially piston driven props, I suspect air temperature, humidity and density have a more significant impact on engine performance rather than aerodynamics... and THAT would be the reason for standard conditions for rating speed and climb rate. Of course, I could be wrong.
RebateMonger
Elite Member
- Dec 24, 2005
- 11,586
- 0
- 0
Cooling of components in a PC will be by conduction (from CPU to heatsink, for instance) and by convection (air blowing past the heatsink).
The humidity of the air will have no effect on conductive cooling. It CAN have an effect on convective cooling.
"Cooling ability" of blowing air will depend on the component's temperature, on the air temperature, on the air velocity, by the flow conditions (laminar flow or not) and by the heat capacity of the air. Humid air has a SLIGHTLY HIGHER heat capacity than does dry air. But not by much. That's why water cooling is more effective than air cooling. If you poured water over a motherboard, it'd cool down faster than if you blew air over it because water has a higher heat capacity than air.
For all practical purposes, I wouldn't expect the relative humdity to have a significant effect on cooling in a PC.
The humidity of the air will have no effect on conductive cooling. It CAN have an effect on convective cooling.
"Cooling ability" of blowing air will depend on the component's temperature, on the air temperature, on the air velocity, by the flow conditions (laminar flow or not) and by the heat capacity of the air. Humid air has a SLIGHTLY HIGHER heat capacity than does dry air. But not by much. That's why water cooling is more effective than air cooling. If you poured water over a motherboard, it'd cool down faster than if you blew air over it because water has a higher heat capacity than air.
For all practical purposes, I wouldn't expect the relative humdity to have a significant effect on cooling in a PC.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 23K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 19K
-
-
-
Facebook
Twitter