All computers produce heat, but computer servers produce a lot of heat – so much that it usually costs a fortune to cool them down. So why isn’t this heat used instead to keep homes or offices warm? Actually, “nerd power” is already being tried out.
Ask Jerry van Waardhuizen about his new radiator and you get an excited response. “I’m very enthusiastic,” he says. “It’s a beautiful thing.” The sleek white box, which has been hugging his wall for two weeks, looks nice enough as radiators go. But what’s really got Waardhuizen excited is what’s going on inside.
Instead of hot water, it contains a computer connected to the internet, doing big sums and kicking out heat in the process. It was created by a Dutch start-up called Nerdalize, and could be part of a solution to a big problem for the tech industry.
We talk about data being “virtual” and stored on a “cloud”. In fact, those clouds take the form of very large, noisy data centres containing tens of thousands of servers. To prevent the server stacks overheating, tech companies spend vast sums on cooling technology – more than a third of a data centre’s hefty energy bill may go on air conditioning. With data centres estimated to account for 1.5% of global electricity consumption (in 2010), this wastage is costly to businesses and to the environment too.
Nerdalize’s solution is, effectively, to spread their data centre across domestic homes linked by fibre-optic cable. The excess heat can then be used instead of going to waste.
The radiators take a little longer than average to heat up – about an hour, Waardhuizen says – and a single unit won’t be enough to heat a room in mid-winter. But, after a small set-up fee, the heat is completely free to users. Nerdalize gets its money for providing data services. During this year-long pilot, its clients include Leiden University Medical Centre, which uses the radiators to crunch through lengthy protein and gene analysis.
The server in the radiator does not stop working when the unit is turned off, but the heat is pushed into an extractor on the outside wall. And in the unlikely event that a user needs heat but the internet is down and the radiator has nothing to work on, it starts performing dummy equations.
Nerdalize have what any start-up needs, after a good idea and financial backing – a nice story. In 2012, Mathijs de Meijer and Boaz Leupe were renovating a run-down house when they accidentally broke the thermostat. “Mathijs was working on his laptop and after a while of really trying everything to stay warm, he said: ‘Well, you know what, how about we take 100 laptops, put them into one of the rooms and then we’ll have nerd-heat heating our house?'” says Florian Schneider, who co-founded the company with the erstwhile housemates. “And that was more of a joke – and then very quickly turned into, ‘Wait, maybe this is not such a stupid idea after all.'”
As they started to look into the idea, they came across a paper published by Microsoft Research and the University of Virginia in 2011 on the potential for “data furnaces” to heat domestic and office properties. The energy savings were so significant, the paper said, that the IT industry could theoretically double in size without increasing its carbon footprint.
The Nerdalize eRadiator
Size: 122cm x 70cm x 17cm – slightly fatter than a normal double radiator
Output: 1000W – about half the peak output of a conventional double radiator, but eRadiators can be left on all day
Cost: 400-500 euros (£290-360) to set up, then free of charge to use
Requirements: A fibre optic connection and an external wall
In fact, the initial impetus for that research was a slightly different problem. Microsoft was wondering what to do with all its old servers, which were less efficient and gave off more heat. The authors noted that the technology required to remotely control and reboot servers had arrived just in time to play a role in cloud computing and “big data”. They envisaged apartment blocks heated by data centres in the basement, and remote “micro-data centres” placed in people’s living rooms.
Nerdalize is one of a number of small companies to have taken these ideas forward, but Microsoft itself has so far put data furnaces on hold.
“After we worked on the project together, I spent about seven months working with Microsoft to try and figure out, you know, ‘Is this going to be the next big thing?'” says Kamin Whitehouse at the University of Virginia, one of the authors of the paper. “And it very well might be a big thing but it doesn’t look like it’s going to be the big thing in the short term.”
Part of the reason, Whitehouse says, is that the big tech companies are currently focused on making billions from the explosion in cloud computing, not saving millions by reselling the waste heat the sector produces.
In addition, there are some limitations. “A network of data furnaces is not equivalent to a data centre,” Whitehouse says. While Nerdalize’s radiators are ideally suited to labour-intensive tasks on a small dataset, like gene-mapping or video game rendering, other jobs require fast processing of large amounts of data.
This is best done by multiple servers working in parallel on different parts of a dataset and talking to one another. Do a web search and a million different computers look at a million parts of the web before the results are collated in the blink of an LED light. The Netherlands boasts a comprehensive fibre-optic network, but the fastest way for computers to work together is still to put them in the same room.
There are also practical obstacles, Whitehouse says, such as the cost of maintaining remote servers, and the issue of data security. He notes that Nerdalize’s radiators are in a tamper-proof case, and use encrypted data but even so, “there are probably a fair number of computing jobs that companies would not push out into people’s homes. They would want to have them in a secure data centre. So again that limits the total market share they could achieve.”
Florian Schneider disagrees. He says Nerdalize’s dispersed network of servers is more secure than a data centre, which represents a single point of failure. “It becomes nearly impossible to know what data is where, when it is there and how to get to it,” he explains.
Click here to continue reading.
SOURCE: BBC News – William Kremer