> I did a small test to see if performance was linear with time. > I wanted to make sure it was suitable for my application. > It seems with both indexed and unindexed tables it doesn't take > significantly longer to do the 1,000,000th insert than it did the > first.
I've discovered there are oodles of factors with this sort of thing, all with depend on what you actually want to do. With 2 million rows, a simple select statement (select * from wibble where key = 1) returning 2000 records out of a dataset of 2 million+, takes between 3 and 10 seconds The *SECOND* time you call this, it's instant due mostly to SQLites caching and HDD caching, however in our case, the 10 second wait at the start was a major issue The only way we could get correct test results for our purposes was to clean boot between every test, and then the results are depressing. Still trying to get past this
