hoffman@hdsvx1.UUCP (Richard Hoffman) (08/28/86)
Robert Pearson writes: > Several people have commented that relational > databases have been improving in performance. I do not know how this has > been accomplished. [He goes on to ask for texts and references on data structures. Can't help you there, Robert, but I can give a little background on the general causes for the improvement:] The makers of RDBMs are getting their act together in a number of ways. The main performance increasers have been: 1) New data structures. For example, Ingres added b-trees to their existing isam and hash strategies about two years ago. That was a big plus for people with large, volatile (subject to frequent update) tables. 2) Better optimizers. There are two aspects to this: o new strategies for looking at a query to determine the best way to evaluate joins on multiple tables, given the available indexes and data structures and the query's search fields, and o new strategies determine whether it's worthwhile to employ these strategies, or whether the query should just be executed dumbly. The first improves the execution of complicated queries; the second improves the execution of simple queries. 3) Better ways to save the query execution schemes for repeated queries. 4) Cleaning up old code. Fixing bugs. Improving the interface. Other miscellaneous stuff that you expect to do as a gargantuan system evolves. -- Richard Hoffman | "Oh life is a wonderful cycle of song, Schlumberger Well Services | A medley of extemporanea. hoffman%hdsvx1@slb-doll.csnet | And Love is a thing that can never go wrong PO Box 2175, Houston, TX 77252 | ... And I am Marie of Roumania." --D. PARKER