Slashdot Mirror
P2P Hard Disk System Warns of Tsunamis
An anonymous reader writes to mention an article on NetworkWorld about a free software application that detects Tsunamis by listening for vibrations in the hard drives of computers. The peer-to-peer network uses the technology that allows HDDs to keep read-write heads on track, and passes the information to a network for analysis. From the article: "If an earthquake that could lead to a tsunami is detected, the supernodes inform the other nodes. Computers running the client software and connected to the peer-to-peer network can then warn of such events. The software is able to provide such warnings because the seismic waves produced by earthquakes travel at about 5,000 kilometers per hour, while tsunamis move much slower at 500 to 1,000 kilometers per hour"
While I agree with you, there is no way to do this in pure java - it will HAVE to have a DLL or some form of native code and it will be highly dependant on hardware. That being said, I agree that it I would be more inclined to run open source code for something like this...
-Em
RelevantElephants: A Somatic WebComic...
In the absense of further knowledge, I'm somewhat skeptical about the hard drives being sensitive enough, but I mean that in the original meaning of "skeptical", as in, updated pending further evidence, not forever committed to not believing in it. Clearly, this guy thinks they are sensitive enough.
But if that hurdle can be cleared, processed correctly the data will be very useful. Most objections Slashdotters are going to raise will be irrelevant. Local aberrations will be cancelled out at the supernode, because the aberrations will only appear at that one node. Simple interference at constant frequencies is also easy to detect and mask out with "Introduction to Signal Processing"-level signal processing.
Merging the data together is a bit more challenging but should be doable.
The only thing I don't see is talking about knowing where the machines are in the real world, which would be very helpful, and that may be coming later. The other thing is that the system probably won't work very well with a simple "IsEarthquake" signal coming out of the clients; the supernodes really ought to examine all the data from its clients and then decide if there's an earthquake. Otherwise, several correctly-timed local abberations could all look like "earthquakes", even with completely different characteristics, if all that is going to the supernode is "IsEarthquake". Of course, the real system may already have both of these things covered and the article merely oversimplified.
Upshot is, signal processing can do some very surprising things with data that seems to consist almost entirely of noise, if you have enough data coming in.
If you read TFA (or even just the summary), you'll notice that it takes quakes of a certain frequency to create a tsunami. High frequency quakes don't cause tsunamis, they just cause a vibration of the water.
Remember kids, tin foil doesn't work, so use LeadHat.
That is the normal range for velocities of propagation of acoustic and elastic waves through rocks. Qualitative table at http://web.ics.purdue.edu/~braile/edumod/waves/Wav eDemo.htm . For a typical cross-section through crust (output from seismic tomography) check http://www.ess.washington.edu/SEIS/PNSN/REPTS/Sum9 7/G03084B.jpg . Look at the color scale on the left for values.