> Donald Griggs and Jay Sprenkle have suggested transactions, which I had not > been using, and I've wrapped the all 800,000 iterations into one tran and > that has reduced the time down to 6 minutes - thanks Donald and Jay!! > Problem fixed - it's now taking as much time to ship the data into memory as > it does to ship it out of Sybase - I'm happy enough about that. I guess it > could be optimised more, but this is ok. > > I've just read your next post about transactions viz-a-viz in-memory db's - > well I see where you're coming from, but the evidence is clearly in favour > of transactions here. My other thought is that opening the db with a > filehandle of ":memory:" is perhaps not 100% correct - that it's somehow > sneakily going off to disk somewhere - can anyone say?
I guess there's more overhead doing individual transactions, even in memory, than DRH thought. It was very interesting to know what was happening under the covers. Thanks Dr H.
