This is pretty neat. The diagram shows a distributed search procedure called nagging. Basically the way it works is that one processor, the master, starts at the beginning and advances along the search. Periodically other processors, naggers, will query the master about the status of the search. The master will tell the nagger how things are going and then the nagger will go off on its own and start running the algorithm an arbitrary number of steps ahead. Once it's made some progress the nagger will report back and then the master will know that it doesn't need to cover the ground that the nagger looked at.

Apparently it works pretty well. All I know is that something about it is very attractive to my inner geek. Also, my inner nerd would like to read a sociology of science paper analyzing the process by which computer scientists came to choose the labels 'Master' and 'Nagger'.

By the by, here's the story of how I came across nagging. The other day I did some fancy stuff over at baseball-reference.com, the result of which was, an automated email from the man behind the site, Sean Forman. The name reminded me of a professor I once had, so I googled "sean forman baseball" and, as luck would have it, one of the top hits was a faculty page at St. Joseph's University, which page had a link to the paper Forman co-authored about nagging. The guy I was thinking of, it turns out, is Sean Foran, not Forman. In figuring that out I found this page and I'll bet that if Bellmaniacs follow the link many will see a face they recognize.

No comments:

Post a Comment

eXTReMe Tracker