Re: Oracle (or Informix, DB2) vs. SAS for Enterprise Warehouse

Jack Sanderson wrote:

> The RDB camp (to which I belong) says that the RDB offers higher
> scalability, better utilization of parallel processors, and much better
> openess to OLAP and CRM tools.

  In my opinion, if you use Oracle (or SQL Server), you will be obliged to persist with external (non-DBMS) tools for the kind of sophisticated analysis that SAS or SPSS provide. But in the longer term it seems to me that such an approach is of limited utility.

  The basic reason for this is that there are significant advantages to integrating your decision support and operational environments. For example, an inventory manager might want to make priority purchasing decisions when stock falls below some critical level. But what level? And how much do you buy?

  SQL gives you arithmetic mean. But answering this kind of question requires statistical algorithms that measure variance (variance, skew, etc), better measures of central tendency (geometric mean, robust analytics), correlation, and some linear algebra stuff for the optimization steps. SQL doesn't do this. SAS and SPSS do.

   In the case of our purchasing manager, realizing that stock levels have fallen more than one deviation away from anticipated values *and* there are no orders in the pipeline indicates that you should cut an order. Quickly. But unless you have a single system with both of these facts in it, making these kinds of decisions is going to be very hard. And modern database applications have things like geography and sophisticated temporal support. SAS and SPSS don't have the kind of generalized query language you would like to support that.

   So there is considerable advantage to pushing these kinds of algorithms into the declarative query framework. Some data management servers do this now. It's only a matter of time before they all do.

    KR

                          Pb