> Why not use a btree index for the >timestamptz column?
There are some capabilities our team lacks. Due to that autovacuum tuning
mechanisms isn't considered at all. It may be in the future.
I know about basic MVCC though. BRIN was an option as the characteristics
you describe match the require
On Sat, Apr 24, 2021, 1:27 AM Mohan Radhakrishnan <
radhakrishnan.mo...@gmail.com> wrote:
> What's your question exactly? If you have confidence that correlation will
> remain high (insert only table, or occasional cluster/repack with cluster
> is done), then BRIN can be a good fit. If you do upda
