ElevateDB 2.03 Build 13 is now available for download. If you're an existing customer, then you should be receiving an email shortly with download instructions. This build contains bug fixes for several incident reports, which are detailed here.
This build corrects one final serious issue with ALTER TABLE that was introduced in Build 10 as part of the changes to improve table alteration performance, as well as a few other fixes. Latest Builds
The last few builds have been coming fast and furious, and I thought I'd take a second to address why this is so. Since January, there have been some major performance improvements introduced into ElevateDB. ElevateDB is starting to mature as a product, and with more and more customers using ElevateDB with very large databases, it was an ideal time to address some performance issues that could be improved upon. These mainly fell into three areas:
Query Optimizer Issues
These issues have been resolved, and most of them had to do with join optimization decisions by the query optimizer, but one early issue had to do with how statistics were computed for the query optimizer. These statistics are used by the query optimizer in I/O cost estimations. The query optimizer is now extremely good at finding the least cost-intensive execution paths for DML statements.
Join Performance Issues
These issues have also been resolved, and were primarily a problem with the internal structures that ElevateDB used to represent filtered sets against large-to-very large tables (2+ million rows). ElevateDB was internally redesigned to use a much more efficient structure for representing filtered sets, and the join performance on large tables has been increased by approximately 40%.
These issues revolve around the performance of importing, exporting, altering, and repairing tables. The importing, exporting, and altering performance issues have all been corrected, with the ALTER TABLE improvements, and subsequent bug fixes, coming in the last builds 10, 11, 12, and 13. The repair functionality will be updated next so that it may be run very quickly, and run online with minimal downtime.
We appreciate everyone's patience while we have worked through some of the issues that have arisen due to these performance improvements. I'm sure that everyone will agree that the performance improvements have been worth it.