I’ve been furiously busy over the past week with the Oracle Exadata announcement. In the very unlikely case you missed it, Oracle introduced a multimillion dollar data warehouse appliance targeted at Teradata and Netezza. I’ve written a lot about Exadata and the Oracle Database Machine over on DBMS2, and the rest of cyberspace has been busy as well. Here’s a guide to some of the coverage.
- Computerworld/IDG’s Eric Lai has an excellent, detailed Exadata overview. Eric also provides links to quite a few other commenters.
- Kevin Closson of Oracle is the most active and followed Oracle employee posting about Exadata.
- I clarified some points in Eric’s piece. I also posted detailed analyses of Oracle query parallelization and Exadata pricing.
- Official Oracle information can be found here and here (although Oracle.com links can be distressingly unstable).
The Cliff Notes (TM) version of all this is:
- Oracle’s high-end data warehousing performance has been disastrously bad.
- The single biggest bottleneck has been Oracle’s reliance on a shared-disk architecture. Exadata solves that.
- Nobody really knows how good the rest of Oracle’s parallel query and parallel analytics technology is, because of how crippled it’s been by the disk access architecture.
- Exadata is a complicated and expensive approach to the problem. Of course, how complicated and expensive it seems to you may be strongly affected by how invested you already are in Oracle licenses and Oracle-savvy personnel.
Oracle’s back in the race and running hard — but it’s still quite a way out of the lead.
