Re: inserting, index and no index - speed
| От | zilch@home.se |
|---|---|
| Тема | Re: inserting, index and no index - speed |
| Дата | |
| Msg-id | 20010611001109.B29684@loony обсуждение исходный текст |
| Ответ на | Re: inserting, index and no index - speed (Tom Lane <tgl@sss.pgh.pa.us>) |
| Ответы |
Re: inserting, index and no index - speed
|
| Список | pgsql-general |
> You might try running the ten thousand inserts as a single transaction > (do "begin" and "end" around them). A HUGE difference (also completely took away the ID field (serial) having only name): Database vacuumed pg: Trying 25000 inserts on index_with... Time taken: 12 seconds Database vacuumed pg: Trying 25000 inserts on index_without... Time taken: 12 seconds <--- MIGHT BE BACUASE OF FSYNC!? (fsyncing also for the previous run) Database vacuumed pg: Trying 30000 inserts on index_with... Time taken: 15 seconds Database vacuumed pg: Trying 30000 inserts on index_without... Time taken: 12 seconds Database vacuumed pg: Trying 35000 inserts on index_with... Time taken: 21 seconds Database vacuumed pg: Trying 35000 inserts on index_without... Time taken: 14 seconds I can't believe what a difference that made. How can it make it faster by putting it in a transaction? I thought that would make it slower. Like only a 100th of the time. Daniel Akerud
В списке pgsql-general по дате отправления: