Slashdot Mirror
Software Telescope
An anonymous reader writes "The BBC News is running the story 'Pyramid power' probes universe which is about LOFAR's software telescope for radio astronomy. The heart of the system is a IBM Blue Gene which processes data from an array of simple pyramidal radio antennae. The array of antennae are also multitasking in the fields of geophysics and agriculture."
Why do projects like this have to be done on supercomputers?
Technical points aside, one political reason for not going to distributed processing is this: some want their data bits to be strictly proprietary.
By distributing, there will be a chance that the distributed data bits would be compromised and captured by someone else (e.g., leading to scooping the cheif investigator). It's a long shot, I know, but that is something that the organization like this need to take into account.
Unfortunately it has also increased (in my perception at least) the number of products that go to market before being well developed and tested.
In the old style development, you had to make sure you had it right before starting production, or else it would cost you a lot of money for retooling and fixing already produced items. Today, you develop a working microcontroller solution (hopefully without hardware bugs), quickly hack together some firmware, and start production and sale.
The consumer will come back with numerous complaints and then is the time to look at the software and maybe release some updates.
Some manufacturers even don't fix software for existing products but just do that in the next model. Early adopters pay the price for a crippled product.