On Wed, Mar 18, 2009 at 06:41:25PM -0400, Michael McCandless wrote: > But at the end of the session you still insist on merging down to one > segment before closing right? So if my added docs all fit in RAM, > closing is fast, but if I go a bit over 1 RAM buffer's worth, then > closing is suddenly slower since it must merge those two runs before > closing?
There's definitely a threshold effect. But the time is still small relative to the total run time of the indexer app. 10 or 15%, I think? > I think having a close that's unexpectedly slow is irksome. It's not unexpectedly slow. It's consistently slow. :D Since KS *always* performs merging of runs at the end, the cost doesn't come as a surprise. Moving all the hard work into a Prepare_Commit() method would solve this and would be a piece of cake to implement. Then we could have a Commit() which was always fast. Marvin Humphrey
