Humidity and server room
I have just had a look at our air conditioning and the environment monitoring in our server room and I have noticed a few things.
Firstly is that the humidity of the room has dropped by 15% from 40% to 25% relative humidity within about 2 weeks. This seems to be since I have turned on one more server. Or is it due to the room having been completely closed for most of that time (ie. people usually go through the room to get to a store room during term times).
The other thing is that using the formula that was posted on the site earlier today, it appears that my servers can produce roughly 40,000 BTU of heat, in a room which is 2.5m x 1.5m in size. The aircon unit is a 3kW wall mounted unit, and has broken down multiple times. So, that formula indicates it is way under spec.
quick back of the envelope stuff indicates that your 3kw ac unit is seriously underpowered for 40k btu.
1 ton of cooling capacity [which i think roughly translates to the capacity of your wall mounted unit] will handle 12k BTU.
The thing to remember is that your equipment isn't drawing the maximum watts during normal operation.
But, even with equipment drawing a proportion of the rated power that would still leave your AC unit underpowered for the area it needs to cool. It could well be that adding that extra server was the straw.
How many servers are we talking about here ? Have you virtualised down as low as you can go, and do you have scope to procure a larger AC unit. Or is it just a case of needing to justify a bigger room and improved ventilation.
We have 9 servers with 700W power supplies, a core switch with a 1200W power supply, and a few other bits and pieces which come to about 2kW total draw. So a total draw is just under 10kW.
Virtualisation... I have made a start, but it is limited in that my request for a) additional hard drives or b) a SAN for the images to go onto were denied.
I'm really just looking for comments, so I can go to my manager and say 'the unit supplied via our LEA is not fit for purpose, so it needs either upgrading, or our usage of servers re-looking at and reducing via a virtualisation project, as mentioned by me earlier in the year'.
I may be able to get rid of around 1400W of power usage in there, but it'll be difficult.
well, the unit supplied doesn't appear to be fit for purpose based on current server count...and more importantly future expansion.
Originally Posted by localzuk
ofcourse now is a good time to talk to management about new servers to replace old. Because it's not efficient to have 9 700w power supplies when a single multicore server could consolidate that right down. the new ibm x3850 m2 for instance, up to 8-cores, up to 8 sff disks, and one or two 835w power supplies. A couple of those could possibly replace your entire server estate.
you've also got to factor in the heat generated by your ups if you haven't already done so. could be another factor in looking at server usage.
Ah, 3 of the servers are terminal servers. (dual quad core xeons with 16GB of ram each, for future expansion to 64bit).
1 of the servers is a iSCSI san box. 1 of the servers is our backup server (ie 2TB storage).
The rest are pretty much virtualisable. I'm looking at virtualising our main AD server at some point this term - which will mean 1 less server. But it comes down to a complete lack of budget as the main problem here.
It's not just your UPS you have to remember about, but also your switches and lighting too.
That's included in the 'other power usage' block i gave earlier. :)
Originally Posted by GrumbleDook