Re: Natural keys vs Aritficial Keys

>> Every so often some crackpot claims to have discovered the perfect hash but bigger keys usually need bigger buckets to reduce collisions.  <<

You might want to look at the research by those "crackpots" that keep getting published in ACM and IEEE journals. Not my idea of crackpots. Just as a starter try "Practical minimal perfect hash functions for large databases" (http://portal.acm.org/citation.cfm? id=129623&dl=).

The bucket size is not the problem. Finding a hashing function that can run fast enough and be minimal and perfect has been the challenge. Cheaper storage and faster processors have made this approach practical. Have you ever worked with Teradata?

>> Same goes for column-based stores, not to mention the trans-relational "model".  Meanwhile, CPU design has been stagnating for years, ... <<

Not a big fan of the multi-core chips that are going to be standard in a few years? Having seen WX2 and SAND, I like the vertical database architectures. I have not played with Vertica yet. It works because it can be compressed and made parallel

I would not fault the CPUs, but the lack of functional programming languages (Erlang, Haskell, F#, etc) for operating systems that could take advantage of the newer hardware's parallelism. Received on Wed May 20 2009 - 22:08:45 CEST