| From: | Alvaro Herrera <alvherre(at)commandprompt(dot)com> |
|---|---|
| To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
| Cc: | pgsql-hackers(at)postgreSQL(dot)org |
| Subject: | Re: Avoiding redundant fetches of btree index metapages |
| Date: | 2006-04-25 17:52:53 |
| Message-ID: | 20060425175253.GC28043@surnet.cl |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Tom Lane wrote:
> Some experimentation over the weekend turned up the result that $SUBJECT
> is a good idea. I had never thought about this much before, figuring
> that in a heavily-used index the metapage would stay in cache anyway and
> so fetching it at the start of each index search isn't costing any extra
> I/O. That's true, but what it does cost is bufmgr contention, and in
> fact more than you'd expect because the amount of work actually done
> before releasing the page again is minuscule. (See off-list discussion
> attached below.)
Wow, this is extremely nice. Congratulations on another well-spotted
performance problem solved.
--
Alvaro Herrera http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tom Lane | 2006-04-25 17:58:27 | Re: Catalog Access (was: [GENERAL] Concurrency problem building indexes) |
| Previous Message | Tom Lane | 2006-04-25 17:43:55 | Avoiding redundant fetches of btree index metapages |