Thanks for all the help and thought to our problem.
Jonah H. Harris wrote:
>
> David has stated that the index to heap visibility check is slowing
> him down, so what are the possible options:
>
> - Visibility in indexes (-hackers archives cover the pros/cons)
> - True organized heaps
> - Block level index (Tom/Simon's earlier discussion)
>
>
Several people seem to have a problem with full index visibility,
however both last thread and this thread people seem to agree that
keeping a page level visibility would be an acceptable compromise.
Perhaps we should implement a rough patch and run some time stats to
post for everyone to weigh the pros and cons against hard figures?
The discussion on block level indexing and true organized heaps, if I am
interpreting this topic correctly, would essentially keep the table
optimized for accessing the data through one index? This seems like a
very interesting idea, but I don't know that it would really solve our
current problem because we need to use more then one index to
efficiently access a table.
We have a very driving need to solve this problem and so don't mind
doing legwork to do feasibility testing or stat collection to help the
community decide the best solution for Postgres.
