Modding Server Rack Back Panel

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

josh

Active Member
Oct 21, 2013
615
190
43
Hi all,

Planning a somewhat ridiculous mod to the server rack as I plan to move everything indoors. My rack's back panel is metal mesh. The airflow with the mesh isn't that good by itself as the rack houses high density units so I'm looking to build an active venting system. At the same time, I'm looking to reduce the noise level as well.

The plan is roughly to:
1. Use mass loaded vinyl to deaden the sound and seal up the mesh holes to better channel airflow
2. Add 8x120mm high CFM fans in zig zag formation
3. Build a small enclosure around the panel and attach high CFM fans at the top to help the hot air move out faster

Visualization sketches to come later, thoughts and criticisms appreciated.
 

Dawg10

Associate
Dec 24, 2016
220
114
43
My first thought is adding a bunch of fans isn't going to help make it quieter.

My second thought is to seal up the mesh, put an exhaust/ 3-speed box fan on top and call it a day.

My third thought is we need to study the problem further.
 

josh

Active Member
Oct 21, 2013
615
190
43
My first thought is adding a bunch of fans isn't going to help make it quieter.

My second thought is to seal up the mesh, put an exhaust/ 3-speed box fan on top and call it a day.

My third thought is we need to study the problem further.
I need some way to evacuate the heat from the bottom of the rack. It's a full rack so a fan box at the top does nothing for the bottom.
Not adding small server fans that need to move large amount of air, will be adding larger fans that can afford lower RPM to move the air. Priority is evacuating air from the server exhaust zone so the intention is to move the air horizontally first before moving it up.
 

Dawg10

Associate
Dec 24, 2016
220
114
43
Re-read what you wrote...

Sentence 1 says: "I need some way to evacuate the heat from the bottom of the rack."

Sentence 4 says: "Priority is evacuating air from the server exhaust zone so the intention is to move the air horizontally first before moving it up."

Perhaps a picture or sketch (or both) would help. I'm thinking you're trying to solve multiple issues with multiple causes?

Also; "It's a full rack so a fan box at the top does nothing for the bottom" isn't necessarily true... Heat removal from the top causes cooler air to enter from somewhere... If that somewhere is the bottom then yay. There are ways to balance air flow to permit zone cooling but, again, pictures and a sketch are needed to move forward.

edit/ I think I see my problem: I read Sentence 1 to mean you which to evacuate air from the bottom of the rack... as in downward...
 

josh

Active Member
Oct 21, 2013
615
190
43
So I made some sketches. Sorry I'm not the best artist out there. :eek:

 

Dawg10

Associate
Dec 24, 2016
220
114
43
Nice work; the drawings get the meaning across so they're fine. Now you need to measure the back-of-rack temperatures from bottom to top; there's an old saying: if you can't measure it you can't control it. Get your hands on a IR thermometer and shoot the hot air exhaust of each rack component and make a spreadsheet or map.

Your idea for mods is sound but needs to be tuned up, although the top vent fan is questionable as, to be effective, it would need to move more air than the 4 rear fans combined. That air is also the lowest density air in the system, thus increasing CFM/ power requirements even more. If you do go with a top fan I'm thinking centrifugal, and it will likely be noisy.

The rack heat map will assist in sizing and placement of the rear exhaust fans. At the moment I'm thinking seal the rear door, replace the 4 small fans with one large air mover exhausting into the new 'add-on' and venting unassisted out the top, but need to understand what the heat profile looks like.
 

josh

Active Member
Oct 21, 2013
615
190
43
Actually, at the moment my rack isn't exactly full so I could theoretically re-rack the units for optimal airflow. Is it better to fill the rack top down or bottom up? Should the high density equipment be lower or the higher than stuff like the switches?

Also, I seem to have forgotten to add the sound deadening materials in the sketches, should I use acoustic foam on the inside of the door, then mass loaded vinyl on the outer layer or just MLV on the inside?
 

Dawg10

Associate
Dec 24, 2016
220
114
43
I was going to ask about how it's populated after you shot the temperatures. In a situation like yours I'd look at how to get the hottest units at/ close to the top. Theory here is the hotter the air/ the lower it's density/ the higher the delta P = higher natural draft = higher velocity up. Free additional heat removal: maximizing thermal updraft by adding heat as it rises.

Acoustics is another ballgame. So far all the changes will only redirect the noise; adding light mass to the back will stop sound traveling through the back, but with only minor absorption. Consider this: if air can escape then so can sound... Houston we have a problem.
 

josh

Active Member
Oct 21, 2013
615
190
43
I was going to ask about how it's populated after you shot the temperatures. In a situation like yours I'd look at how to get the hottest units at/ close to the top. Theory here is the hotter the air/ the lower it's density/ the higher the delta P = higher natural draft = higher velocity up. Free additional heat removal: maximizing thermal updraft by adding heat as it rises.

Acoustics is another ballgame. So far all the changes will only redirect the noise; adding light mass to the back will stop sound traveling through the back, but with only minor absorption. Consider this: if air can escape then so can sound... Houston we have a problem.
My present rack is configured with the heaviest and largest units at the bottom and I'm starting to see how this may not be the most optimal. Will do proper planning this time before rerack as the units are quite heavy and inconvenient to move around.

I'm also slightly unsure what to do with the switches that don't extend all the way across the rack. I'm currently mounting them from the back but that leaves a large hole in the front which I've covered with blanking panels. However, bit concerned that this means the switches have no intake of cold air and might be recirculating the warm air from the back?

I actually have mass already added to the front and it really helps with killing the sound from the intake fans which are the loudest. I'm trying to recreate the Netshelter CX, by forcing the air to take a longer route, hopefully dispersing the sound as it travels to exhaust.

Not sure if I should use dense wood or metal for the added layer?
 

Dawg10

Associate
Dec 24, 2016
220
114
43
Wood or metal is likely going to depend upon ease of installation, but the greater the mass the better. Metal has the disadvantage of being more reflective but that can be mitigated.

It sounds like you have a real good idea re: the physics and thermodynamics involved; I'd suggest you get a better idea of what heat is being generated where and how it's moving through the rack, along with uncontrolled cool air infiltration that may be fighting you. Once you understand the nature of the problem the answer becomes clearer; next steps involve listing & applying practical remedies. I draw/ write all this shit down; something about black & white...

Moving a bottom mounted UPS/ stepdown transformer often isn't practical; mine, for example, weigh in at 120 and 95 lb. respectively. The obvious other concern is lead-acid batteries leaking but a drip tray and regular inspection routine would cancel that out. Isolating and removing UPS exhaust would remove it from the rack heat profile altogether.

BTW; I have 3 racks in service, including a 18U CX. At some point I intend to do a photo essay discussing how the APC design works, along with a few ways to optimize further ie installing temperature controlled variable speed exhaust fans in place of the stock single-speed units.
 

josh

Active Member
Oct 21, 2013
615
190
43
Just a quick question, what are the temps like in the CX with high density cloud servers like C6220?