Actually, I’m not at the office, but I am doing what I would consider “work” today. I’m at the TDWI World Conference in Orlando.
Today I attended Stephen Brobst’s class on “Designing a High Performance Data Warehouse”. While the course came highly recommended, I was somewhat skeptical going in that, being software agnostic, it would be too theoretical to really put to use. But I was astounded by how much content was in this course.
We reviewed:
- Join Strategies
- Indexing Strategies
- Database Parallelism
- Partitioning Strategies
- OLAP & Aggregation
There was a whole lot of stuff in between. I’m going to need to bone up on employing partitions and parallelism both in the databases (Oracle and SQL Server) and in our ETL tools (Informatica and eventually SSIS). There is a lot that can be done architecturally that could help our processes scale. He had a particularly good illustration Amdahl’s Law, demonstrating the benefits of parallelism, that I will need to draw out and put up here. The point being that many processes need to be run sequentially, but you get a huge boost out of finding ones that can run in parallel and doing so.
I will hopefully have more details later, but I need to run to a session that Informatica is putting on. I will probably be coding tonight. Put some of this creative energy to work.