02-24-2012 03:33 AM
A dataset is not more or less a physical SAS table.
A data mart is rather a concept. A Data mart can consist of one or many (then linked together by keys) tables, typically made for ad-hoc reporting, and source for cubes and information marts (which is data prepared for a more specific purpose).
02-28-2012 01:33 PM
In a data mart you can have many data sets.
A data mart is a place you can storage your data sets.
The data sets are the place you storage your datas like your informations.
In fact is more than that, but you can start thinking like this.
03-07-2012 09:57 PM
But could we design a datamart just based on SAS Datasets? Unique Keys can still be defined in SAS datasets.
In my workplace, we use Teradata as our datamart because it it much faster. We also use in-database processing options. One of the benefits cited for in-database processing was the calcuation is done in Teradata and only the results are sent through the network to the sas client. Whereas without indatabase processing, the full data will have to be first downloaded from Teadata datamart and then processed on the SAS Client. But if the SAS datasets resides on the SAS Workspace server, can the speed equal Teradata processing?
03-08-2012 03:04 PM
As stated earlier a datamart is a concept and can be implemented in many ways. Keys are not integral to the concept. The data could even be in text files. When you add the additional structure, you are adding functionality that can make the work more productive.
As far as whether a SAS Workspace server can equal a Teradata system for processing speed, nobody can tell you. There are so many variables that impact processing speed that I am sure you can come up with one where Teradata is better and another where the SAS Workspace server is better.