Обсуждение: Variation between query runtimes

Hi,

I have the following quering plans:

"Seq Scan on ind_uni_100 (cost=0.00..27242.00 rows=1000000 width=104) (actual time=0.272..2444.667 rows=1000000 loops=1)"
"Total runtime: 4229.449 ms"

and

"Bitmap Heap Scan on ind_uni_100  (cost=314.00..18181.00 rows=50000 width=104) (actual time=74.106..585.368 rows=49758 loops=1)"
"  Recheck Cond: (b = 1)"
"  ->  Bitmap Index Scan on index_b_ind_uni_100  (cost= 0.00..314.00 rows=50000 width=0) (actual time=61.814..61.814 rows=49758 loops=1)"
"        Index Cond: (b = 1)"
"Total runtime: 638.787 ms"

from pg_stast_get_blocks_fetched i can see that both queries need almost the same number of disk fetches which is quite reasonable ( the index is unclustered).

But as you can see there is a great variation between query runtimes.Can someone explain this differnce?

Thanks!

""John Top-k apad"" <topkapad@gmail.com> wrote
>
> from pg_stast_get_blocks_fetched i can see that both queries need almost
the
> same number of disk fetches which is quite reasonable ( the index is
> unclustered).
>
> But as you can see there is a great variation between query
> runtimes.Cansomeone explain this differnce?
>

Can you give a self-contained example (including what you did to clear the
file system cache (maybe unmount?) to *not* let the 2nd query to use the
file content from the 1st query)?

Regards,
Qingqing