Ideally, administering a relational database management system should be simple — describe the tables, load the data, and rely on the system to take care of everything else. Complexity comes primarily in two (somewhat overlapping) forms:
- Manual steps required for the system’s regular operation, that in principle could be automated away, but actually haven’t been.
- Manual steps need to tune the system for performance.
Vendors whose products shine in one of those areas but not in both tend to claim greater advantages in “simplicity” than they actually possess. And the list of such vendors is long, because there’s something of a negative correlation between excellence in the two metrics, often because:
- Older products tend to require more tuning, but tend to have more mature automation tools.
- Newer products often need less tuning, but might not yet have all their tools up to snuff.
Vendors of older products of course dispute this generalization, at least in their own specific cases. “Yes, we have great tuning options, but you don’t HAVE to use them. Our out-of-the-box — or at least lightly tuned — performance is just as good as the other guys’.” But I don’t know of anybody who thinks it’s best practice to set up an Oracle data warehouse (Exadata or otherwise) on a load-and-go basis. More to the point, I don’t know of anybody who thinks that — porting of existing applications perhaps aside — it’s as easy to get an Oracle system ready for high-performing production as it is one from Netezza, Greenplum, or Kognitio. And the same applies so Sybase IQ, DB2, and so on.
Vendors of newer products dispute this generalization too. “Look at our live customers, running large databases, with minimal amounts of DBA effort.” They have a point. On the other hand, I don’t know of any Netezza, Vertica, or Aster Data customers running databases as complex as those often found on Teradata or Oracle.
If you’re buying data warehouse technology — software, appliance, or service alike — you probably expect your data warehousing needs to shift and grow rapidly over the years ahead. Projecting the effort needed to keep your data warehouse abreast of the change isn’t easy — but to some level of approximation it has to be done. And to a not inconsiderable extent, the same goes for more “routine” administrative burdens as well.