Sun to Create Underground Japanese Datacenter

Kurtz'sKompund writes with word of a Sun project in Japan, one that's taking a somewhat non-standard approach to data center construction. To save on power, heating, and water costs, the consortium is going to be building their center in an abandoned coal mine. The outpost will be created by lowering Blackbox systems into the ground; estimates on savings run to $9 million annually in electricity alone.

Are they crush proof?

[rueger]

The Blackbox containers are robust enough to withstand earthquakes, being capable of withstanding a quake of magnitude 6.7 on the Richter scale.

I don't know, but placing servers 100m underground in a place that routinely is hit by large earthquakes seems a dubious idea. The containers themselves may survive a quake, but what happens when the disused coal mine collapses onto and around them? Even if the containers and servers survive, will the power and data cables? If the tunnels collapse how will you get to and from the servers for maintenance?

Re:Thermal fun

[RallyNick]

The temperature in a cave means nothing, unless you take into account the cave's ability to dissipate heat somewhere (water or air moving through the cave). If you go inside a cave that's been at constant 55F for a thousand years and you suddenly heat it with 50 kilowatts of power from your data center the temperature will settle at 255F in a hurry. About the only advantage you get from a cave is a constant supply of really cold water (if sufficient rain that year). Ambient air temperature is irrelevant since usually you don't have a strong draft in a deep cave and static air will heat up pretty quickly.

Re:Savings in Electricity...

[Anonymous Coward]

According to TFA, $9M could be saved on electricity when using 30'000 server cores. Also according to TFA, 10'000 cores are planned with a $405M budget. If power demand scales directly with the number of cores, this would equate savings of $3M annually. Based only on these savings (which of course won't be the only factor, but since TFS and TFA single them out so clearly), this project breaks even after a measly 135 years or about five and a half times Sun's current age.

You need to rethink your analysis - your calculation assumes the cost of a comparable conventional datacenter is zero.

You need to subtract the cost of a comparable conventional datacenter from $405M, then divide by your $3M annual savings in cooling costs to see when you break even.

(I'm sure other operating costs will vary as well)

Re:air humidity

[WoLpH]

Since the computers produce a lot of heat the humidity wouldn't be much of a problem, try putting a computer in a humid garage, the computer will be just about the only thing dry in there.

I do wonder how much this thing will really save, I wouldn't be so surprised if the costs are comparable to the normal installation (remember, the normal installation costs for these things is near 0, they just need a power, network and water plug). If they'd just put the server somewhere with some other cooling source available (a lake for example?) it would probably be even cheaper.