blogs.conchango.com

I think you can replace Teradata with a large number of database vendors and in particular the new boys on the block i.e. the appliances such Netezza. Also this approach puts more strain on the BI platform and I suspect will begin to play more and more into the hands of the heavy weights such as MicroStrategy.

"Please Dan please tell us about the good old Express days"

Well not quite, however when Oracle bought (IRI) Express they decided to reengineer it and draw the OLAP engine back into the relational database. In theory less data needed to be moved into a seperate data store, or held twice (and if it did at least it was "close" and well integrated). So will Oracle do the same second time around with Essbase?

"cube [...] object type inside the database engine"

An OLAP cube is an N-dimensional data structure - in other words an analogue of a relational table, which is also an N-dimensional data structure. The real question therefore is why do we need two different models to achieve the same thing? The answer has more to do with the limitations of the SQL language than with the relational model or relational engines.

Continuing with the idea of the cube as a separate object from relational tables due to the limitations of the SQL language, the only question is about the advantages of a file system repository for storing data and metadata instead of a relational one, specially at the time of quering.

And then we can think about use of the other resources such as memory and cpu. I really don't know enough to assess the pros and cons of each approach.

In terms of processing, a closer metadata relationship between relational tables and multidimensional ones could provide better (even native) ways of loading data to cubes.

Well! I have a 1 year experience of Teradata and before that i use MS SQL server. Basically Teradata (TD) targeting only big customer and what the client want is the expensive product which realize everyone (comparators, inside office,  gain market people confident ) that we have barrier to enter in a market. TD offers them hardware of 1 million $ approx and customer believe is, if he spend huge amount then there is no need for any thing else, u may say they have purchased a best practices for life. Where as with Microsoft and Oracle we can do the similar things by using a strategic and intelligent approach like purchase the best H/W. Follow all the recommended steps given with these products u will get a same system with less TCO and it will give u more ROI. One more interesting thing the professional of TD are less and costly also where as with MS & Oracle we have many professional in market.

But there is no doubt that TD performs very fast because of its approach which is discussed in article.

So if i conclude my opinion then this comparison/decision is more lie on the management or investor side rather than on technical side