So, it’s been awhile. Surely, you think, I come bearing some wonderful story of our times here. A trip to the country? How we spent Easter? Maybe a treatise on the language? Well I am sorry to report that I have little to report. A rather nasty head cold laid me low for a bit of the break, and the remainder was spent discovering that no matter how big of a holiday Easter may be in the smaller towns, it is merely an inconvenience in the city. The stores and commercial establishments (excepting restaurants, bars, and clubs) are closed Sunday and Monday. Other than this, any festivities that went on must have happened in residential areas or some such. Most of my weekend was spent wandering the streets and parks of the city, occasionally stopping to write a bit or read some such. Margit Island remains a wonderful place, but since I’ve already devoted an entry to that I shall refrain from repeating myself.
Instead, I shall explain what our bloody project is about. Since it is currently consuming far too many waking moments I suppose it has earned a place in this blog anyway. Now you’ll first have to know a bit about grid systems. In a nutshell, a grid is a heterogeneous (i.e. made of different types, say a PC, a Mac, and a supercomputer for example) network devoted to distributed processing. Many very long, hard problems in the scientific world can be broken into smaller, more manageable problems. These can then be solved on their own, and recombined to find the solution much faster than would otherwise be possible. The trick, you see, is to take all these fragments and solve them at the same time. Grids make this, theoretically, easy. You submit the problem, a software agent called the resource broker allocates some computers to work on it, and off it goes. Then they churn merrily away, and some time later your answer pops out.
Our specific problem is to analyze the performance of an economic marketplace model on a grid, versus on a single computer. Since one algorithm processes many sets of data, it should be easy to parallelize the system and dramatically speed things up. Unfortunately, there are a great deal of pratfalls and unexpected delays in the real world. So extensive testing is needed to see if it is practical, as opposed to theoretically nice. At this point I’d rather not get into how the project goes…suffice to say that the ideal world of perfect grid systems will happen around the same time that world peace is achieved and Dick Cheney stops eating babies.
No cabbages!??! ..but cabbage is so delicious!
ReplyDelete