Application Tips for Mini Splits in Server Rooms

[Marc O'Brien]

People, if you don't yet subscribe to the UK's RAC magazine then here's a good reason why you should. They publish my articles in the associated RAC Service Engineer Magazine. Of course this argument can be substantiated - everyone should read my articles and we all doubt anyone would disagree.

Thoughts on better application of mini splits in server rooms. It has only been out a few days but already I've received a hand full of phone calls from suppliers and contractors wanting to discuss the principles further.

The link is to a low resolution scan of the most recent issue's article.

Let me know what you think.

[JimmyG]

What people?? This place is dead with just some poor looser wannabe's hangin... LOL LOL LOL

[Joseph]

Interesting idea Marc, congrats on being published. I have been having trouble getting a couple of server rooms to keep up with the growing loads being added to the racks (not large enclosed racks though). I am wondering do think this principle could be applied to an open rack sever room? Also would some conversion give a strained system a little more capacity? Funny thing is even though the loads have gone up the icing issues haven't disappeared.

Joseph

[r.bartlett]

It's trying to get people to get away from thinking that server rooms must be kept at a constant 18 deg'c throughout the room. The idea of the chiller strips is to stop the hot air from mixing with the room air before it returns to the conditioner. It's best to think of the warm air out side of the server racks as the start of the return duct and design from there.

Correctly ducted the 'air on' will be 30+deg'c and therefore will hold more heat to help prevent coil freezing and save energy by not allowing it to mix with air that's just been cooled. This method will also increase the performance of the existing equipment by supplying cool air to where it's best needed -i.e. the air intake side of the servers and not where it's not.

The big problem is converting conventional rooms to this configuration....

Personally I would have liked the r/h diag to have the return in the centre and the cold to the outside. (as per chiller strip picture) Then draw the chiller strips in for a better explaination of the design. However I know that the articles are sub'ed heavily and time isn't always available..

Cheers Richard

[Joseph]

be kept at a constant 18 deg'c throughout the room.

Most of the server rooms I work on are set at 68/69*F. The couple rooms that are having trouble (icing minisplits) are set at 71*F.

Last week I was called out to an iced unit. 3 or 4 ceiling tiles had been taken out and not replaced. Unit head was 12" or from the open attic, replaced tiles, unit hasn't re-iced.

Most of the server fans I see blow out the back of the rack. My problem will be airflow from the minisplit heads that are wall mount, not ducted. Small server rooms, with growing heatloads. I have the go ahead to change one out to catch up with demand. Have to give it some thought, might be worth looking at for an upgrade.

Joseph

[New Directions In Building Ser]

Don't really like minisplits protecting Electronic Equipment (incl LAN) Rooms as the quality of the units is usually not reliable enough and if someone's such a cheapskate that they'd use a minisplit on a 100% sensible load then they're probably too cheap to build in any redundancy. Unless these minisplits are high quality and high airchange rate units.

You want to stop icing, vapour seal the room. I mean Joseph - what's this about ceiling tiles. Glue the bastards up seal all the holes (and get some of that stuff that gives birds a hotfoot to identify and inhibit the culprits), put double strip rubber seals on the doors and paint every surface of the fabric with acrylic. then no ice. Anybody that wants to stuff things up... just set the room setpoint below 20 degrees (you increase the risk of drying out the room as the coil strips more moisture over time and elevate humidifier running, you also increase the risk of short cycling and increase the risk of failure).

Given all that Marc's idea does have merit for some applications (cheapskate customers) but as for the pomy "Service Engineer" subscription, I think that I prefer to invest my subscription money in the Chartered Instutute of Building Services Engineers Journal (another pomy publication) and our local Australian Institute of Heating & Refrigeration Engineers Journal, the Institute of Plant Engineers of Australasia Journal and Climate Control News.

[r.bartlett]

The hot isle cold isle server farm room solution is a well researched idea, so I'm surprised you haven't heard of it.?

So actually it's not the a cheapskate option but is a much better system to ensure correct airflow through the server cabinet. The fact that it leads to a better utilisation of the cooling duty (and hence lower energy costs) should be aplauded too.

Cheers

Richard

P.S 'Service Engineer' is a 4 x per year free issue from the people who produce R&A.C magazine.

[Marc O'Brien]

[Marc O'Brien]

My idea behind drawing the cold isle in the centre was so that the hot isles we then exposed to the outdoors or surrounds so that in winter they are cooled by cooler walls and in summer there is less heat transmission through the walls on account of a reduced TD across the walls.

[Abby Normal]

I am still stubborn, I vote for high airflow and an oversized fan coil