The extra processing overhead (even if it's in another thread) simply takes too long.
The problem is that the error checking is done _regardless_ of whether or not there is an actual error in the data, if the conditions are set on the table itself.
What you might find is: the larger the data set, the less likely you are to have "time" to execute error checking routines, or foreign keys for that matter.
By the way, the reason (I think) B-Eye has restricted the interface, is because I was getting 12 to 15 "spam comment posts" per every half hour (about 1.5 years ago), so they had to shut the comments down.
I'll let them know that it's challenging to post comments, but I do welcome them.... Genesee (I'll be posting GENERAL content courses shortly).
I had some experience with the interesting Solaris OS, which is a 64-bit OS but run a lot of 32-bit applications.
One thing good about Solaris is that it allows the 32-bit application to allocate 4G heap memory instead of 2G (or 3G) than Windows, Linux and AIX.I'm about to extend my course-ware to "generic ETL/ELT" courses which go through each point in detail, as to why and where, and how....But I do believe in the sharing of knowledge, so look for more information to be posted shortly.There frequently is one (maybe two) architectures that work a super-high volumes...So not to be a broken record, but I'd like to know - when any of you "test", please disclose the row size in bytes, and the number of rows, and the conditions of the target table under which the architecture was tested.Thanks, Honey If you search through my blog archives for Benchmarks you will see that Kettle does not do very well on performance tests against other ETL tools.