What's Next, Massively Parallel BerkelyDB?

All of Monash's comments about BrightHouse lately forced me to go check out their web site.
Said to be a column-oriented data warehouse based on MySQL, I was both intrigued and horrified - column stores are great, but MySQL? Really?

Their web site, and the one available white paper, aren't exactly information dense, but they're enough to give some sense for the product. It's column orientated and features compression as well as some very interesting index-like structures. In general, it feels like a pretty powerful if comfortingly simple package.

Architecturally, it's essentially a MySQL storage engine plus some other components, built into or on top of MySQL, depending on how you look at it. This is pretty smart, on one level, and is essentially the same approach taken by Dataupia - build a plug-in to an existing database to eliminate the need for new interfaces, SQL dialects, etc., thus making it easier to sell and easier for customers to implement. The downside, though, is that it ties you to someone else's product plans. More importantly in this case, however, it ties you to someone else's image - Dataupia chose Oracle, InfoBright (makers of Brighthouse) chose MySQL.

Really? MySQL? Does anyone with any real experience with MySQL want to build and manage a 10TB data warehouse with anything based on MySQL, even if you put someone else's engine under the hood? No one I know does, that's for sure. And I sure as hell don't.

I have to admit that there are plenty of people out there using MySQL though, so maybe InfoBright's plan is not to compete with other data warehousing products, but simply to provide a growth path for existing MySQL customers. That approach makes a lot of sense to me, actually, from a business perspective. From a technology perspective, however, I think I'd rather have... well, just about anything else.