Solar-powered data centre won't get IBM very far...

Too Little Information

This summary article is almost a carbon copy of IBM's press release [at http://www-03.ibm.com/press/us/en/pressrelease/35891.wss -- thanks for including that by the way] with just a sprinkling of assumption thrown in between the loose facts of the original and used to support an unproven conclusion.

Here are the two key assumptions (granted, made because IBM's press release is very vague) as I see them:

Assumption 1: the data center uses the entire 50kW of power supplied by the solar array. IBM's press release never mentions how much power the data center actually draws (only that it runs off of the solar array and that the solar array provides 50kW), but makes a big deal of mininizing AC-DC/DC-AC conversions and other power conservation methods. So benchmarking based on a traditional data center would be useless. In order to draw a rational conclusion as to the viability of this system, this would need to be cleared up because:

Assumption 2: the solar system does not store any excess power. This, too, is not mentioned in the press release from IBM. But if the system did store excess power, it will be able to run for more than 20% of the time -- how much depends on the actual answer to assumption 1 and the efficiency and capacity of the storage system.

Misleading...

"That means the sun is not powering the data centre for over 80 per cent of the year:"

That means the Sun is not powering **100%** of the center 80% of the year. But from dawn to dusk you'd actually have a sort of bell-curve of power production as the solar component ramps up towards 100% while the grid ramps down to 0%, flattens out when the sun reaches optimal position, and then down again on the other side as the solar array ramps down and the grid draw goes up.

The way the article is written implies that if the sun is providing anything less than 100% of the power used that this is somehow a failure. Frankly even if the array provides only 30% of the power on a given day, that's 30% less energy used and while perhaps not cost effective it is reducing power demand for the facility.

Seems like a good idea

Presumably as server hardware costs a fraction of the power it takes over the servers lifetime you can do a literal "follow the sun" virtualisation scheme with several data centres around the world. All the racks have to have UPS battery backup anyway so short periods of no sun aren't a problem. Add in a local swimming pool or town to take the heat extracted by the AC systems and you are onto a really good thing. I'm not sure exactly how much heat a fully populated blade enclosure chucks out but I suspect it is enough to supply hot water to several houses.