Some folks at Microsoft Research and the Computer Science Department at the University of Virginia have tabled an idea that is both radical and utilitarian at the same time.
The idea is centered around so-called, Data Furnaces, basically, servers, or clusters of servers, that will act as heaters inside homes or buildings.
The researchers note that: “phyiscally, a computer server is a metal box that converts electricity into heat. The temperature of the exhaust air (usually around 40-50 degrees celcius) is too low to regenerate electricity efficiently, but is perfect for heating purposes, including home/building space heating, cloth dryers, water heaters, and agriculture.”
“We propose to replace electric resistive heating elements with silicon heating elements, thereby reducing societal energy footprint by using electricity for heating to also perform computation,” a research paper called, “The Data Furnace: Heating Up with Cloud Computing,” noted.
There are a couple of dimensions to the idea. First of all is cloud computing and fibre networking, which now enables computer resources to be virtualised, so it is feasible at least to place all these servers remotely inside someone’s home or a commercial facility. The second is the fact that existing data centres generate a huge amount of heat, so they need lots of energy to keep them cool. By adopting a distributed computing architecture, it also distributes the heat, hence not only making that heat available to locations where it might actually be needed, but also reducing the cost of operating data centres.
In a way, the idea is not so much an alternative form of heating, but rather a method to reduce the being “wasted” inside current data centres today. And it isn’t that new an idea. I know of at least one trial by IBM in Europe, which takes the hot water generated by a data centre’s water cooling system to heat a nearby swimming pool.
The new study, however, has done a little more qualitative analysts into the concept. Since IT is estimated to consume 3% of the total US electricity load per year, the report argues that, since home heating already constitutes about 6% of the total US energy usage, by piggy-backing on only half of this energy, the IT industry could double in size without increasing its carbon footprint or its load on the power grid and generation systems.
The authors of the research are proposing three different types of DFs.
A low-cost model featuring older and less energy efficiency (hence hotter) servers and disks for installation in residential homes that piggyback on the users existing broadband connection. While these type of hardware don’t work efficiently in data centres, they are “ideal for providing heat and maintaining basic disk and network operations year-round,” and “supporting delay tolerant batch jobs,” the authors argued.
Another type is so-called, low-bandwidth neighbourhood DFs, which still use existing broadband connections but leverage their proximity to end-users to compensate for the limited bandwidth. “Network latency can be reduced by creating more data centres closer to customers, particularly for cloud applications such as E-mail, multimedia distribution, and low-hit rate Web pages and services that are delay tolerant, embarrassingly distributable, and/or can benefit from large-scale replication and geo-distribution,” the authors wrote.
Perhaps the most viable immediately are eco-friendly urban DFs for commercial complexes, which will be supported by dedicated fibre or leased lines. This type of DF would allow data centre operators to expand into urban areas without having to lease sites, but do present the challenge of having to dissipate heat during summer months, when the heat is not needed. Also, the distributed architecture introduces a different granularity to perform massively distributed computations, like MapReduce, the report said.
The idea has some obvious challenges, as highlighted by the researchers – for example, making sure that DFs don’t break usage policies of existing broadband lines, or ensuring the physical security necessary for data centres. They also have to enable zero-touch management or the maintenance costs would far outweigh any other benefits.
The different deployment models certainly need a lot more work to find the balance between its benefits in providing heat, and its ability to support data processing for operators and end-users, but the concept does present some intriguing possibilities. There really is no reason why server manufacturers and service providers cannot get together and produce a home-type server that would relief some of the load of today’s massively centralised data centres – imagine an Apple TV-like device with server-level processing and storage. And there’s no reason why such a device can’t be integrated with a home heating system.
At the very least, someone should find a way to recycle the heat from existing data centres, especially those inside commercial buildings, into some usable form.
You can download the report here.