B+tree discussions

[Vincent Ladeuil]

>>>>> "robert" == Robert Collins <robertc at robertcollins.net> writes:

    robert> At this point, I think we have a 'good enough to be an improvement'
    robert> candidate, once some remote testing is done. (And I'm on that now).

    robert> B+Tree, no bloom, with 1000 page cache outperforms or is equal with
    robert> GraphIndex on every benchmark I've come up with. I think in real world
    robert> larger-still indices it should be tolerable if not better due to lower
    robert> memory consumption, and it should be tunable (simply by upping the cache
    robert> size) to deal with such cache thrashing workloads.

    robert> I think bloom use is a compelling future feature, but its not a clear
    robert> enough win /today/ and we can take the time to make it better later.

    robert> If everyone agrees with this concept, I'll make the stackable formats
    robert> about to be merged, also include this index layer.

+1

If adding bloom later implies a format bump then:

- +0.8 for adding it later,
- +1.2 for adding it now.

   Vincent