The U.S. EPA estimated that servers and data centers were responsible for up to 1.5 percent of the total U.S. electricity consumption, or roughly 0.5 percent of U.S. greenhouse gas emissions, in 2007. With companies such as Apple and Google strongly pushing the move to cloud computing, that figure is likely to increase significantly in the coming decade. Since a lot of energy is consumed keeping the computer systems cool, colder climates are seen as more favorable sites for data centers. But a new paper from Microsoft Research proposes a different approach that would see servers, dubbed Data Furnaces, distributed to office buildings and homes where they would act as a primary heat source.
The Microsoft Research paper says that at around 40-50°C (104-122 °F), the temperature of the exhaust air from a computer server is too low to regenerate electricity efficiently. However, this temperature is perfect for heating purposes, such as home/building space heating, clothes dryers and water heaters. So the researchers argue that placing servers used for cloud computing operations directly into homes and/or office buildings would turn heat generation from a problem into an advantage.
The Data Furnaces (DFs) would be micro-datacenters on the order of 40 to 400 CPUs that would be connected to the Internet and integrated into the house/office building's heating system in the same way as a conventional electrical furnace. By leveraging the home's existing infrastructure and doing away with the need for dedicated real estate and construction of new facilities, DFs would significantly reduce the cost per server when compared to conventional data centers.
Additionally, such a setup would also provide lower network latency as the storage and computation systems can be located closer to areas of high population density and therefore those using them.
The DFs would be managed remotely and the researchers suggest that cloud computing service operators could provide free heat to host families in return for occasionally replacing air filters or, in extreme circumstances, turning servers off and on.
With residential areas much less physically secure than a dedicated data center facility, the researchers say each DF should have an individual tamper-proofing device, such as a networked sensor, and all stored data and network traffic must be encrypted. Software running on the servers would also need to be sandboxed and secured from the hosting party.
The researchers say replacing electric resistive elements with silicon heating elements would allow a smaller energy footprint by using electricity used for heating to also carry out computation. By piggybacking on just half of the six percent of U.S. energy consumption used for home heating alone, the researchers say the IT industry could double in size without increasing its carbon footprint or its load on the power grid and generation systems.
Even taking into account the generally higher cost of electricity in residential compared to industrial areas, calculations detailed in the paper indicate substantial financial savings. Examining a number of different system structures, the researchers say that when compared to the US$400 cost per year for a server in a conventional data center, the estimated savings per DF per year range from $280 to $324.
So although the thought of Data Furnaces being placed in people's homes may seem unlikely now, with computers and network connectivity continuing to get faster and cheaper, and energy costs continuing to rise, the idea might not be so far fetched.