Nucleus Database from SAND

[OLAPer]

Hi people!

I am starting a new role where the company has implemented SANDs Nucleus database. I was wondering whether anyone reading this might be able to give me an insight into the workings of Nucleus.

Most importantly the database uses Bit-Map Vector Indexing, which I believe is not really indexing, it is implied in the data structure (??). Would appreciate any information on this.

Also I got the distinct impression from my interviewers that there is no need for "Cubes", they further suggested discontinuing the use of Cognos. The question is, can a database with a very fine grain, really perform as well as a Cube in Cognos???

Any assistance, views and feelings on the points above would be really appreciated.

Kindest regards


OLAPer

:

 

[jnicho02]

I used to work with the guys who wrote Nucleus.

Basic principle involves turning any field into a one-bit answer, e.g. sex M/F/U turns into three fields, sex-M sex-F and sex-U meaning that you can use the fastest index (a bitmap) and create combination queries quickly.
Range values will need to be banded. I believe there is a tool for deciding on band sizes.

There are then some highly cunning structural tricks that make it really quick.

The general answer is yes, this can work considerably faster than other technologies. The work is done in the pre-processing. Jeremy Nicholson, Director of

http://www.firstlightassociates.co.uk

a UK-based Java and Data Warehousing consultancy

 

[jnicho02]

You peaked my interest, so I took afurther look.

http://www.sand.com/library/whitepaper_exploration.pdf

describes it better. Think 'data compression' like pkzip. Jeremy Nicholson, Director of

http://www.firstlightassociates.co.uk

a UK-based Java and Data Warehousing consultancy