Engineering A Data-Driven Infrastructure
Corsair's Publishing

Good article, but you’re being too nice. And I say that in response to an entry that talks about the quantity of data molestation.

Item #1: Don’t just instrument it, parameterize it. Everything you do should be a lesson in how to do it better next time. So many databases fail because they are hyper-tuned to the task at hand, which makes it impossible for them to adapt when the business changes (usually 48 hours after go-live).

Item #2: If transformation is a noticeable part of your storage process, you’re doing it wrong. Seriously. Go home and think about what you’ve done.

Item #3: Yes, the analytic database should be a mirror of your production data- to a point. It’s a dark mirror. This is where transformation is appropriate, but again: if your transformation is anything more than OLTP to OLAP optimization (showing my age here), you’ve already lost.

Item #4: I can use it. The fact that this makes me stand out in a crowd is sad, but you started the self-promotion. I can walk through any schema. Most people can’t. You should be designing for them.

I’m not even going into the quality issues right now. Good points, right down the line. If anyone actually paid attention to them, it’d be a lot harder to find work in Business Intelligence.