Re: The Nelson-Pick Data Model

"Laconic2" <laconic2_at_comcast.net> wrote in message news:4pudnZrZrem7PBPcRVn-tg_at_comcast.com...
>
> "Steve Lancour" <stevel_at_lancour.com> wrote in message
> news:2I2dnRkcubpcnRDcRVn-2g_at_comcast.com...
>
>
> > Raining Data (the successor to Pick Systems), offers this overview of
> > the model: http://www.rainingdata.com/products/dbms/d3/d3datamodel.html
>
> Thanks, Steve.
>
> I folowed up on the Raining Data link. After looking at the picture of a
> data cube, and the brief description of the Nelson-Pick model, it
reminds
> me very much of the "Dimensional Model" that Kimball writes about, in his
> books on Data Warehouses and Data Marts.

While the model permits OLAP sans materialized data cubes by an ease of making virtual data cubes, the model is really not all that similar to MOLAP, ROLAP, etc in that sums are not (typically) stored, for example.

The ease in making virtual stars is due to the ease of making virtual fields that look to the user as if they are the same as the stored data. So, if you take a "relation" table in the Chen sense and then take each attribute in the key (each dimension) and use it as a foreign key to the dimension tables/files, you have a single "entity" that appears to be a cube, or star join. The sums are then also specified as derived data, rather than stored.

While I would not classify Pick in the same category as MOLAP tools, it is often used instead of such tools simply because there is no need to use ETL tools to load a data mart or warehouse if you can use a virtual cube. Even if the data should be frozen, simply copying the original files and using virtual cubes is often as effective as anything else PLUS you can run any production reports against the frozen data since it stays in its original format.

There are tools in the OLAP environment that run with a Pick engine, such as Ascential's (Ardent) Data Stage product, which has IBM's UniVerse as its engine. However, even though there is marketing literature that tries to put the multidimensional spin on Pick, it really is dissimilar in many respects. I think I used too many words to say that, but I hope it made a little sense.

> It sounds as though D3 would provide many of the same features as Business
> Objects or Cognos Power Play. Is this the case? Also, how easy is it to
> retrieve data, from one's programming language of choice?

Data retrieval is typically done with the multivalue query language or DataBASIC, but one can use Visual Basic, .NET languages, and Java as well with most Pick providers. One can also use SQL, but my experience says that is a means of using the worst in both SQL and the Nelson-Pick model to use them together. The interface from languages other than the two native languages (query language and DataBASIC) is not standardized across database vendors in the multivalue world.

> Thanks again.

Sorry to pipe up on this one as I'm sure Steve could clarify, but since I've done quite a bit with OLAP, data marts, cubes, star joins, and have Kimball's books on my shelf, I figured I'd chime in. Cheers! --dawn Received on Mon Nov 08 2004 - 00:21:28 CET