The Data Furnace

In joint work with Microsoft researchers, we propose installing servers into homes, apartment buildings, and office buildings, and to use the exhaust heat as a primary heat source for the building. We call this the Data Furnace. This approach has several advantages: 1) it precludes the need for a dedicated heating system, which currently accounts for 40-60% of a typical building's energy usage 2) it lowers the cost of operating a data center by about $300 per server, or 75% of annual amortized costs 3) it moves data and computation closer to the user, providing lower-latency service because your favorite websites are hosted on the local area network of your building (1 millisecond away instead of 100 milliseconds away). Since publication of the paper, several people and companies have taken to the idea for personal or professional prototypes, and one user is posting data live to the web.

Some people are already trying to reuse the waste heat (see articles below) from data centers that are being built anyway, for other reasons. The data furnace takes this idea one step further: the opportunity to use waste heat is enough reason to install new computers into a building. Doing so precludes the need to (a) build expensive new data centers, and (b) spend energy solely for heating the building. In current work, we are creating new Computer Science techniques to turn this idea into reality.

Publications

Jie Liu, Michel Goraczko, Sean James, Christian Belady, Jiakang Lu, Kamin Whitehouse. The Data Furnace: Heating Up with Cloud Computing. The 3rd USENIX Workshop on Hot Topics in Cloud Computing (HotCloud'11). Jun 14-15, 2011. Portland, OR.