Re: Is Optimizer smart enough?
| От | Tom Lane |
|---|---|
| Тема | Re: Is Optimizer smart enough? |
| Дата | |
| Msg-id | 8514.1136956125@sss.pgh.pa.us обсуждение исходный текст |
| Ответ на | Is Optimizer smart enough? (tmorelli@tmorelli.com.br) |
| Список | pgsql-hackers |
tmorelli@tmorelli.com.br writes:
> Just for curiosity: suppose there is an excellent index frequently picked by
> the optimizer. Suppose now that this index became extremelly fragmented with
> thousands of updates. Without a REINDEX, will Optimizer still pick it? Or the
> optimizer is smart enough to detect index fragmentation and discard it?
IIRC, btcostestimate is sensitive to the physical size of the index,
so it would catch the first-order effect of index bloat. It wouldn't
notice index fragmentation in the sense of increasingly random location
of logically-sequential leaf pages. But I'm not sure how much that
matters for typical situations. If you have a huge fraction of dead
tuples, I'd say that index fragmentation is not your worst problem...
regards, tom lane
В списке pgsql-hackers по дате отправления: