p-v-o-s / pioneer-valley-open-science.github.com

9 stars 2 forks source link

DIY home data furnace #47

Open sekjal opened 12 years ago

sekjal commented 12 years ago

Summary

Provide a model for building low-cost hardware that generates significant enough heat to warm a human home. Hardware could be recycled from older models found in first-world countries. Computation done could either be for personal usage, or sold on a market to offset electricity costs.

References

Microsoft research paper on the subject

Hardware

Either using recycled machines or low-cost new equipment. Care would need to be taken with the overall efficiency, as heat excessive heat is generally a bad thing for computers, but good for people.

Aesthetics would also need to be considered; many people won't want an old grey or tan CPU sitting in their living room. Cheap and easy case mods that make the overall package more visually pleasant (or dual-purpose as some kind of furniture) could be published to an open repository, or done as artistic works and sold by craftspeople.

Computation and market

By networking these machines, and registering to some kind of dispatcher, a market could be created to sell CPU cycles to bidders. Keeping the machines running at the appropriate percentage of capacity in order to maintain the desired temperature would require careful calibration and rationing of computation. If a simple conversion formula could be developed mapping work to heat, all the better.

The inevitable result of this market would be "computation season" in the Fall and Winter, when folks are looking to heat their homes. With international adoption, this could be evened out some, but probably not completely, as there are fewer potential users in the Southern hemisphere than the Northern, and Spring/Fall would likely see less demand than Summer/Winter.

Thermostatic feedback

Connecting the data furnaces in a home to a series of temperature sensors would be necessary in order to provide the appropriate feedback for calibration and allocation of work. An arbitrary number of sensors should be linkable to an arbitrary number of furnaces, under the control of a single program.

sekjal commented 11 years ago

Rather than selling computation itself, why not sell Cloud Server space? Networking a bunch of old hardware together in a redundant array could be done in such a way as to make a local cloud server farm, good for serving up local content, or caching frequently accessed content in the region so as to lessen overall network load.

Introducing the "Internet of Things", additional seasonal heating 'fuel' could be added by monitoring real world conditions locally, without the need to sell processing power to others (at least as much).