Chapter 9. Operational Issues and More

After the base database objects have been created and the initial data has been loaded, you have a candidate production data warehouse. You might begin with just a few facts and some simple aggregates, and you might have just a few hundred million rows to start, but the general concepts of how to successfully deploy and manage that data warehouse will remain exactly the same as you scale from these very humble beginnings into a full-fledged, multi-terabyte behemoth. Always bear in mind just how big you think the data warehouse will be 12 months into the future when making any operational support decisions. If you don't, then you will most definitely want to keep your resume up-to-date for when you run into the proverbial brick wall of problems as you scale above and beyond a terabyte.

The first and key thing to remember is that a data warehouse is not your traditional OLTP database. The deployment and management of the data warehouse must be treated very differently. You will generally find that much, if not most or all, of your traditional DBA bag of tricks will not be advisable or even feasible. You must very quickly learn to think well outside the box and openly embrace radically new and often unorthodox techniques, including those clearly outside the traditional Oracle DBA toolset. You also need to realize that good advice and techniques for data warehousing may make little or no sense in the OLTP world. So do not too quickly judge an idea as poor if it makes no sense. For example, OLTP DBAs would never fully index tables, but data warehouse DBAs must. The more you can let go, the more likely you are to succeed. It reminds me of the original Star Trek episode where Captain Kirk and crew are in the Old West and must relive the gunfight at the O.K. Corral. To survive the gunfight, they must fully disbelieve everything their senses tell them is real?anything less and they're dead. This is good advice for the aspiring data warehousing DBA.

In this final chapter, I'll present some thoughts to help you think outside the box. But I cannot fully detail any of these issues since much will depend on your customer's needs, database size, database version, operating system, hardware, and many other issues. My goal is hopefully to expand your horizons regarding the possibilities and inspire you to think well beyond the obvious or traditional solutions.