Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!ncrlnk!ciss!dbruck From: dbruck@ciss.Dayton.NCR.COM (Don Bruck@ciss.Dayton.NCR.COM) Newsgroups: comp.databases Subject: Re: Recent postings to comp.databases... Message-ID: <876@ciss.Dayton.NCR.COM> Date: 4 Aug 89 21:03:56 GMT References: <15963@pasteur.Berkeley.EDU> <450@cimshop.UUCP> Reply-To: dbruck@ciss.UUCP (Don Bruck@ciss.Dayton.NCR.COM) Organization: NCR Corp. Network Application Services Lines: 36 In article <450@cimshop.UUCP> davidm@cimshop.UUCP (David Masterson) writes: >> >Would you say, though, that it is more appropriate to start with a database >that well constructed (ie. normalized) and work toward an efficient design or >vice versa? This, of course, assumes that a person is willing to "wreck" his >well built database for the lowly consideration of performance. ;-) > This is why many people suggest the separation of data design responsibility into two. The first does analysis of the data: understand what it means to the user and what the keys (data, relationships) are. (That to me is what normalization is all about, defining the base elements of data.) The second person is responsible for taking the normalized data and making it useful, i.e. de-normalize it, add indices, foreign keys, etc. This person must understand the problems of denormalization and explain to users (including programmers) that the database will now perform better *BUT* it will not be as flexible. This gets us back to a question related to the original: Why go to all the trouble of Normalizing if someone is just going to denormalize? Because when you go through the whole process you can measure how much you have "hurt" your original. If you don't go through the process of finding perfection first then you can't measure how good the system is. You also don't know which trade-off you should truly use and which access will need to suffer since you cannot optimize all accesses. -------------------------------------------------------------------------- Don Bruck NCR Corp. Corporate Data Planning and Administration PCD 6 1700 S. Patterson Blvd. Dayton OH 45479 My opinions and interests are my own.