Re: If it moves, we can track it!

From: Eugen Leitl (eugen@leitl.org)
Date: Thu Oct 17 2002 - 05:31:54 MDT


On Wed, 16 Oct 2002, Lee Daniel Crocker wrote:

> Let's see...there are about 6 gigapeople on the planet, and it
> takes less than 32 bits the store a location on the Earth's
> surface to around meter resolution, and you'd probably want

Hey, you're not tracking a race of teleporters. Movement is always
continuous, incremental, few m/s on foot or a bit faster in cars. Anything
larger/faster (bus, train, ship, plane) will be tracked as a single unit.
And current location services have orders of magnitude less resolution, so
you need less bits to represent.

> updates at around .1-second resolution, so the total possible

Much too quick. Even 1 Hz refresh rate I mentioned is overkill. 0.1 Hz
would be more than adequate, and I would adaptively go slower for slower
objects.

> raw incoming data stream would be on the order of 200 Gb/sec.

Actually, if you just want to track all objects without any analysis or
recording, and use clever coding, a single box with several GBytes RAM and
GBit Ethernet would do the trick. Cute, but useless, so you still need a
cluster for real work.

> Conveniently, most of those 6 gigapeople are stationary for
> long periods of time, and they tend to cluster themselves
> neatly into a few thousand clumps, so data compression would
> probably get you down into the 100 Mb/sec range, which seems
> well within the bounds of even present technology.

Absolutely. A small cluster of consumer grade boxes would do. If it wasn't
for the cost of distributed smart sensors, it would have been done
already.



This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:17:38 MST