Re: Insert speed question

Hello Alvaro,

El 02/06/2004 9:58 AM, Alvaro Herrera en su mensaje escribio:
> So fix your query!  Also what do you expect to happen if you put
> constants in the column list?  This certainly looks like a mistake to
> me.  Anyway you should really format your query better so you can
> understand it and see obvious mistakes.

I'm sorry I did copy a wrong piece of the clipboard :(

>>                                                        QUERY PLAN
>>

>>---------------------------------------------------------------------------------------------------------------------------
>> Seq Scan on hisventa  (cost=0.00..633607.24 rows=4882546 width=149)
>>(actual time=26.647..363517.935 rows=4882546 loops=1)
>> Total runtime: 1042927.167 ms
>>(2 rows)
>
>
> So you are inserting 4 million rows.  This makes a lot of I/O so no
> wonder it takes a long time.  I'm not sure if the time is reasonable or
> not though; 4M rows/1M ms = 4 rows/ms.  Not that bad.

Agree, insert time got better

>>>I would ask the question otherway round. What is the level of performance
>>>you are looking at for your current workload. By how much this performance
>>>is worse than your expectation?
>>
>>Since I have not tested the server with the production workload yet,
>>maybe my perpception of performance is not rigth focused, basically my
>>expectation is database must be faster than the current old legacy
>>Foxpro system.
>
>
> If you are going to have big load, you should at least try to code a
> simulation with big load, doing random queries (not any query but the
> actual queries you'll get from your system -- for example if this is a
> web-based app you can try to use Siege or something along those lines).
>

I'll take your word and will make such tests



--
Sinceramente,
Josué Maldonado.
"Tiene algo que ocultar aquel que se toma a mal las críticas." Helmut
Schmidt. Político alemán.