Re: Forecasting with RAM/Express and general advantages of ROLAP approach to data warehousing

Hi Richard,

You can do forecasting and what if analysis with RAM. Like the MOLAP environment, you can reference the RAM formulas in your forecasting method. All you have to do is add your forecast/what-if measures to the RAM application once the build is complete. This enables you to treat the RAM database like any other Express database. I have clients who use both programs and/or EIF files to create these definitions.

It is important to remember, that RAM is built using Express, and any Express command that can be applied to a normal MOLAP Express database can be applied to a RAM one. There are only two areas to be careful of. Firstly, if the dimensionality of the forecast/what-if cube is different to that of the RAM cube, then you will need to set the Measure dependency property of the forecast/what-if cube to point to the RAM cube. This is not difficult and can be done through administrator. Secondly, if you create a cube that uses the MCALC command, then you may run in to issues. This is because RAM uses MCALC to support the Sales Analyzer custom aggregates function.

The only real down side of the ROLAP approach with Express is that the system is obviously slower. This reflects the need of the RAM data model to get the data each time a user queries a variable. However, with the caching model, you can tune and store data with a high hit rate. Secondly, the complexity of the system may appear greater at first, but once you understand how RAM works, it is very simple.

RAM has improved dramatically over the last few years and is a great tool for deploying star schema based data.

I hope this helps.

Mathew

richard.campbell_at_esa.int (Richard Campbell) wrote in message news:<25c7d0cf.0109270522.1045e5fa_at_posting.google.com>...
> What-if Analysis using databases created by RAM
>
> If I understand correctly, in order to allow what-if analysis, Express
> uses variables that can be modified, which once modified are reflected
> by changes in custom measures or formula.
>
> RAM maps/converts RDBMS facts to formula and not variables, does this
> mean that forecasting or what-if analysis cannot be performed?
>
> From the broader perspective I am trying to understand the advantages
> of using a MOLAP tool to those of using a straight
> datawarehousing/RDBMS approach. Initially I imagined three advantages
> of the MOLAP approach, Forecasting (what-if analysis), speed and
> improved analytical functionality.
>
> The impact of the latter is lessened by the improvement of analytical
> function built-ins/extentions to SQL. If what-if analysis is not
> available I can see
> minimal benefits in using a ROLAP approach to a simple data warehouse.
> (A guess speed is only a product of summarisation and reduction of
> volumes of data)
>
> However, I am just beginning to use Express and therefore the
> possibility that I am overlooking something important is a high one -
> any enlightenment to additional advantages will be gratefully received
> (Especially with reference to Oracle product offerings)
>
> Thanks
>
> Richard
Received on Fri Sep 28 2001 - 16:36:24 CEST