DSXchange

What is typical in terms of refresh frequency of the hierarchies in the dimension server? Are there often automated daily ETL loads? If much of the enterprises's BI required daily refreshes then can we assume the dimension server needs a daily refresh?

Also, how do automated ETL loads created by techies keep from interfering with manual maintenance that may be done by business users through the Razza GUI? and vice versa

Your question about refresh frequency also makes me wonder about implementing slowly changing dimensions in the dimension server.

Thoughts on the best practices approach to the good old friend of data warehouse developers?

Typically the dimension server controls the reporting hierarchies for the enterprise as a whole, so BI processes get approved organization, account and other hierarchies from the dimension server. The hierarchies, or reporting structures, are managed independently from application data by the data stewards for each dimension, who control how frequently those dimensions change.

This is a business process issue more so than a technology issue.

After rereading your question, often if the dimension server is not the source for new nodes in a particular dimension, like accounts, then ETL feeds may occur daily.

There are multiple ways to get data into the dimension server, all of which are subject to the same rules and validations that user maintenance are subject to, so integrity is preserved. It is very often the case that ETL loads into the dimension server occur much more frequently that data is extracted.

datastage wrote:Your question about refresh frequency also makes me wonder about implementing slowly changing dimensions in the dimension server.

Thoughts on the best practices approach to the good old friend of data warehouse developers?

The dimension server does an excellent job of supporting slowly changing dimensions because it can provide hierarchies from any point in time and can provide surrogate keys in many cases.

It should be noted that many clients archive the feeds they receive from RazzaDS rather than pulling historical hierarchies each time period. Either method will work, though.