In the technique of aperture synthesis, multiple small sensors have their signals combined so as to form, virtually, a sensor the size of the outline of the whole grid with the gaps filled in.
The effect is to make a virtual telescope that can see much more than you'd think.
In optical or radio astronomy, this matters because the smallest object that can be resolved is inversely related to the diameter of the telescope. So, by linking a small telescope in the USA with one in Sweden, effectively a 5000 kilometer diameter "dish" is created, virtually. This is a standard production technique.
The key is not only that the signals be brought together, but that the phase information has to not be lost as well. This is subtle but crucial. (See VLBI information on wikipedia for the idea.)
A mathematically similar type of thing occurs when multiple databases in different places are made to "act as one", which requires each of them to give up a little bit of control in order to establish a realm where "two-phase commit" works. The concept is that when an update shows up at any site, none of them should recognize it as "true" until they are all ready to recognize it as true, at which point they all change as one, so there is never a time at which some databases think the fact is true, and others think it is false. This is the key difference between an actual distributed database architecture and a tinker-toy grid of legacy systems that send messages back and forth to each other, with each one retaining its own authority to decide for itself whether something is true or not.
The cost of not yielding control is that the "aperture synthesis" part goes away. Suddenly, we're back to having 14 small dishes that talk and argue, not one virtual huge dish that sees things as one that none could see separately.
The point is that it cannot work if it is compartmentalized.
Compartmentalization and emergent aperture synthesis are incompatible.
If some unit wants to be sure it is the only one that ever sees the aggregate whole, then the cost is that the aggregate whole will be tiny and flawed. Either everyone sees the new truth, or no one sees it.
So, organizational politics of top-down control and secrecy interferes, fatally, with abilty of disparate units to act as one. And, no, a top-down tree-structure doesn't fix this problem.
Assuming the above analysis is correct, sooner or later someone will accomplish just exactly such a synthesis. Effectively, the IQ of their organization as a whole will jump by a factor of 100 over everyone else. Self-interst would argue that the someone should be us.
First side to overcome theory X with theory Y wins all the marbles. The clock is running.
(Image of the Very Large Array, as featured in the movie Contact, is from wikipedia.)
technorati tags:theoryY, aperture, synthesis, aperturesynthesis, telescopes, collaboration, fusion, phaselock, phaselockloops, PLL, control, organiztions
No comments:
Post a Comment