On Thu, 2005-12-01 at 19:00 -0800, Mark wrote: > So followup question: are any of these other systems good > with large'ish databases? I have multi-TB Oracle databases that > I'd like to explore on a "free" database. Is that realistic? > My current system relies heavily on partitions, bitmap indexes, > table compression, though I can live without query rewrite. > > For example, I have a ~10 billion row, ~1TB table, time partitioned, > with about 10-20 million rows per day. I need "interactive" > performance (2-10 seconds) response on queries that are confined > to 2-3 days data. Obviously, I can't scan a TB table each time. > In Oracle, the I/O pruning is simple to setup, and then automatic, > with partitioning. > > Are there similar facilities in MySQL, or the others?
That's a slightly larger data set than anything I've worked with, but if you don't need the materialised views with query rewrite then I'd say it's worth investigating the latest Bizgres to see how far it's got, then relaying your experience back via the Bizgres mailing list. It's got a few of those handy OLAP support facilities like a loader GUI, etc. http://www.bizgres.org/ http://pgfoundry.org/projects/bizgres/ Currently Bizgres has no table compression, but on the scale of things that is an "easy" feature to add, so I wouldn't be surprised if it comes soon. Perhaps it would cost less than a year of an Oracle license to implement ;). Sam.
