Relational theory requires you to abstract the database to fit a (badly flawed) prescriptive theory of what data *should* be like. MV simply models the database to look like the real world.
Relational scatters the data about individual real-world items across multiple tables. MV (properly designed) puts it all in one place. That's why MV doesn't need optimisers etc - it doesn't need to guess what data is likely to be accessed as a blob - if data is tightly joined in the real world, it is likely to be tightly joined in an MV database (while it's guaranteed to be scattered everywhere in a relational database). Relational REQUIRES that data comes in two dimensions. MV reflects the real world in accepting data that comes in more than two dimensions. A generic "data is n-dimensional" theory will ALWAYS be preferable to a "data is 2-dimensional" theory. I liken that requirement of relational to Euclid's assertion in geometry that parallel lines never meet. Without realising it, he placed an artificial restraint on geometry and held it back centuries. C&D have done the same for database theory. Would you use Newtonian mechanics to model events in the LHC? The physicists would laugh you out of Switzerland. As David said, Oracle market to CEOs - relational wins on marketing clout. Unfortunately, the mathematical foundation behind it is iredeemably flawed :-( It breaks the Einsteinian corollory to Occam - "make it as simple as possible - but no simpler" - relational theory has been simplified too far, with the result that it ends up far more complex than it need be. Cheers, Wol -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of David Jordan Sent: 13 August 2008 01:18 To: u2-users@listserver.u2ug.org Subject: RE: [U2] Why Buy (or develop in) UniVerse? Hi Louie Intersystems have done some benchmarks of multidimensional databases versus RDBMS and some of that logic follows through to UniVerse. It is difficult to compare UniVerse to RDBMS in benchmarks as they are designed for RDBMS strengths. If a benchmark was designed for UniVerse strengths instead, RDBMS would not look so rosy. RDBMS databases are designed to optimise cache and indexing because of the performance issues in the database. UniVerse does not inherit those performance issues, hence they do not need to optimise Cache and indexs to the same extent and the optimisations needs to be different. There are hosts of differences. RDBMS have fixed length and fixed structure records, where as UniVerse has variable length records and fields can be added at any time. More UniVerse records can fit on a disk sector than RDBMS rows increasing U2 performance. RDBMS don't efficiently lock rows, they do group locks. Universe can lock individual records without performance hits. RDBMS work with optimistic locking as pessimistic locking is a nightmare with group locking. UniVerse can handle both optimistic and pessimistic locking. RDBMS stores all tables within one file, UniVerse has a file for every table. Totally different approaches for BU, Restore and handling file corruptions. RDBMS have to join multiple tables which creates overhead and referential integrity issues. UniVerse stores all specific data in a multidimensional record. UniVerse is close to Zero-Administration, where RDBMS still require expensive Database administrators. RDBMS have large workloads in setting up security access to tables for different users. UniVerse can use table security or OS file security. RDBMS have limited functionality in business rules stored in the database. UniVerse can handle complex business rules with ease. In complex applications UniVerse is well ahead. However the argument should not be technical. The CEO and board does not make decisions on Cache and indexes, they make it on a business case. ROI, Cost of running, Staff numbers to administer and develop, competitive advantage. The old joke was what hardware does Oracle run best on, a projector. Oracle markets to CEOs and does little technology discussion, that is why they are successful. Its horses for courses, but if a project is going to be complex, the success rate of the project completing on time and on cost in UniVerse is near 100%, on an RDBMS the numbers are scary. Regards David Jordan Managing Consultant ------- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/ ------- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
