Next Generation Data Centre for High Density Server Clusters

By | 1st July 2015

Firstly, I am no architect. Secondly something like this has been done in Norway. This is just an idea I have had bouncing around for a while and I have decided to let it free.

I had a few design criteria for this idea:

  1. In South Africa we are running out of power generating capacity and having rolling blackouts. The current power generation for the type of data influx we are going to have is vastly inadequate. So power generation is going to be a huge problem.
  2. Having the data centres as close to the end user is still a critical design point, you need the data (or at least the Content Delivery Nodes) as close to the end user as possible. They still need to be roughly pleasing to the eye.
  3. They have to be very high capacity (1200 sqm+)

So, you take a standard cooling tower for a coal fired power station. You pull out all the current structure inside and build 20-30 floors of datacentre space in rings round inside of the tower. (approximately 100 sqm per floor). For added renewable energy you put a wind turbine on the top to recover some power from the wind forced up the tower.


Here are the advantages of the system:

  • Low terrestrial space usage – you are using 3D space instead of flat floor space
  • High power recovery using the wind turbine (this will be more efficient in cooler weather)
  • You can add solar generation to the outside for power generation on sunny days when the turbine will not be at its peak performance
  • Heat generation required for the venturi effect is created as a by product of the server load
  • You can make them pretty.
  • Solar power could match server load characteristics if you are using “follow the sun” deployments.
  • The hotter you make the tower (from a load perspective) the more heat is generated, the more heat, the more wind, the more wind, the more cooling.

Disadvantages of the design:

  • Precipitation – this is a biggie, wet servers don’t work so well. If this cannot be resolved structurally, then a dry place is required. Drier places will be less preferable because you are relying solely on environmental cooling.
  • Possible requirement for a new server design. We are so stuck to what a server must look like, in the coming age we may need to rethink it.
  • Construction – if you don’t have a few unused towers lying around this is going to be an expensive solution.
  • Server access will require qualified mountain climbers with danger pay.
  • Servers would be running at environmental temperatures so all hardware would need to be newer as it is designed to run a bit hotter than older technology.


  • Sorry guys, I didn’t do the math, you would need to work out the power generation by the turbine, solar panels and grid power (I am assuming you would still need grid power)
  • at around 1000 sqm of servers (30 servers per sqm, 1kW per server) you are consuming 30000 kW of power or generating 30000 kW of heat, is that sufficient to drive the cooling tower?