| From: | Michael Paquier <michael(dot)paquier(at)gmail(dot)com> |
|---|---|
| To: | pinker <pinker(at)onet(dot)eu> |
| Cc: | PostgreSQL mailing lists <pgsql-general(at)postgresql(dot)org> |
| Subject: | Re: Index loading methods |
| Date: | 2017-03-27 13:06:55 |
| Message-ID: | CAB7nPqSVpFyHmkx8zx1DxLy8sryYKtH_emtkp=YWapOS41wHTw@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-general |
On Mon, Mar 27, 2017 at 8:36 PM, pinker <pinker(at)onet(dot)eu> wrote:
> If PostgreSQL decides to use an index, does he every time load the whole
> B-tree into memory? or maybe loads only specific subtree or some chunks of
> index?
src/backend/access/nbtree/README provides details about the algorithm
of Lehman and Yao used in btree indexes. In short, backends share
buffer pages, and those will stay around if they are hot enough. If
there is a high page eviction for those btree pages, you may finish by
needing to reload the tree, but with a proper tuning (enough memory)
that won't happen.
--
Michael
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tom Lane | 2017-03-27 13:46:08 | Re: migration to 9.6 array_accum memory issues |
| Previous Message | pinker | 2017-03-27 11:36:28 | Index loading methods |