A conference attendance that tops 2000 suggests that a technology involves a certain number of subtle angles. MySQL became a hit because installing it and manipulating tables were so simple–and yet when you get serious, the simple things start growing hair.
Take replication as just one example. Amost everybody uses it, because who wants to be left without data in case a server goes down? Yet there are innumerable gotchas, most of them inherent in the operation. These gotchas range from replicating a statement that inserts a timestamp or host name (it can produce different results on the slave) to slowdowns due to inequitable resources (a master can run many threads to update the database, but replication is serialized). See our MySQL in a Nutshell and High Performance MySQL for details.
The attendees I’ve talked to at this conference have their hands greasy every day dealing with problems of growth and performance. They’re here for practical solutions, and the sessions I went to seemed to deliver. There’s little talk of new storage engines such as Falcon or PBXT, but a lot of heavy-duty coverage of InnoDB’s internals and behavior, and a few sessions on MySQL Cluster.
In the opening keynotes, MySQL CEO Mårten Mickos and Sun CEO Jonathan Schwartz extolled open source. Schwartz’s talk could almost function as a college classroom introduction to the value of open source, covering a range of standard observations. Both emphasized their support for Linux as the foundation of the LAMP stack, which has concerned some observers. For instance, Schwartz said, “We understand that the stack is LAMP, and not WAMP or anything else ending in AMP.”
This does not keep MySQL from declaring that one of its major goals, post-purchase, is to optimize the database engine for numerous operating systems and hardware.
Mickos also showed off the new data schema design tool, MySQL Workbench, not mentioning that it’s currently available only for Windows. (I’m sure that will change any day.)
Amazon CTO Werner Vogels followed with a keynote promoting S3 and EC2. He placed them in a historical context, positioning them as the newest natural stage in an evolution that began with basic sorts of virtualization, such as separating business logic from the database. Along the way, he quoted the notorious advice from 37Signals not to try to build systems that scale. I don’t believe anybody at this conference takes that seriously.
Here are a few more sessions and BOFs I went to:
- Baron Schwartz (who won MySQL’s annual award for best code contributions) gave an entire session on the EXPLAIN statement, which offers lots of optimization hints to the discerning disciple but requires substantial knowledge of database engine internals.
- Sheeri Kritzer Cabral (who won another MySQL annual award, this one for community) tossed out a fast-paced list of cool tips for avoiding work–both human labor and database load. These are not optimizations in the traditional sense, but things that make you slap your forehead and say, “What a smart way to get where I want to go faster.” (Slides are available online.)
- Chuck Bell showed off MySQL online backup.
- Sphinx was present at the show and got some attention as a text search engine that complements a relational database.
Eight sessions were going all the time, and nearly all were stuffed; people often packed the back walls after the chairs were filled. The energy at this conference is unusually high.