Tuesday, August 28, 2007

Jim Gray and real computing

I used to have a bookmark of Jim Gray's piece "Rules of Thumb in Data Engineering". I have no idea which old profile the bookmark is hidden under, but googling "'sequential access' 'disk' 'rules of thumb'" brought it up as the first hit. Another win for Google over bookmarks (though I was a bit lucky to remember the phrase "rules of thumb").

The piece examines the changing ratios among CPU, disk capacity, network speed and other basic parameters and reaches some interesting conclusions about databases and caching. It also makes the larger point is that with Moore's law and its cousins in effect, the basic assumptions behind our engineering trade-offs change faster than we think. The article itself was written in 1999. The trends it outlines look to be holding up well and the main point even more so.

As you may know, Jim Gray was lost at sea last January. Gray's body of work is classic engineering -- developing a firm, objective grasp of the basic facts in order to put all this wonderful machinery to good use. Even a quick look at his home page shows what a loss this is for our profession, to say nothing of those close to him.

No comments: