Is An Historical Database Always a Data Warehouse?

[ngarris]

Sorry, I posted this in the Resource forum but that was incorrect.

Although I've been in Information Technology for decades, I have no experience (yet) with data warehousing.

We have many operational databases. We have developed a need to store static, historical data. Over the last decade we have designed a few historical databases by adding date columns to an operational-type relational database. And now I'm reading about the dimensional data warehouse and the Kimball methodology.

Are historical databases generally used for the same purposes as data warehouses? Or are there purposes for which a non-dimensional historical database would be a better choice than a data warehouse? Basically we'd like to keep some of our data around for a longer time than our operational databases provide for.
1. Some of our data needs to be retained because it provides a history of our important master tables. Or we may want to refer back and understand something that took place in, say, 1998.
2. Other data needs to be retained because we may want to re-use it. These are text documents which are analyses of complex issues and if the issue is revisited we'd like to take advantage of the previous analysis.
3. Still other historical data is used for ad hoc queries and analysis.

Do you recommend that all of these be supported by data warehouse(s) together with an ETL system(s)?

[ngalemmo]

There are various schools of thought on this. Some would argue that the primary purpose of a data warehouse is as a historical repository and integration point of business data. Those who subscribe to that typically implement somewhat normalized databases.

A dimensional data warehouse (i.e. dimensional modeling techniques in general) is primarily geared towards collecting data for the purpose of analysis, paticularly large ad-hoc queries against large data sets.

Either method can be used successfully to satisfy historical data requirements, but a dimensional model is much better at supporting analytics.

But creating a data warehouse is not a technical solution to a technical problem. It is a technical solution to a BUSINESS problem. I would not create a data warehouse if all that is needed is to roll old data off into an archive. Successful data warehouses are long-term programs involving significant support and funding from the business. It is something the business needs to initiate (maybe with some suggestion), sponsor and support.

[ngarris]

That's helpful. Thanks!

[BoxesAndLines]

Missing from your set of requirements is integration. If I do not need to integrate data from across the enterprise then there is little justification for a data warehouse. The data warehouse breaks down the silos of information stored in redundant applications that do not share information. When the business asks how many widgets did I sell last month or how many new customers did I add last quarter and the answer takes 2 weeks of data extracts, compilation, consolidation, and reporting, people start recognizing the value of a data warehouse.

[Sponsored content]