Re: Insert Rate/Second

• Michael McMullen <ganstadba_at_hotmail.com> wrote:

> This is good for one time inserts. It becomes less
> accurate once the query
> gets into the shared pool, but you can still use it
> as a base.
> select substr(sql_text,14,instr(sql_text,'(')-16)
> table_name,rows_processed,
>
> round((sysdate-to_date(first_load_time,'yyyy-mm-dd
> hh24:mi:ss'))*24*60,1)
> minutes,
>
>

trunc(rows_processed/((sysdate-to_date(first_load_time,'yyyy-mm-dd
> hh24:mi:ss'))*24*60)) rows_per_min
>
> from v$sqlarea where sql_text like 'INSERT%'
>
> and command_type = 2 and open_versions > 0
>

Thanks Michael. This looks good.

This has confused me somewhat though. I would have thought that there would be a 1-to-1 ratio between executions-to-rows processed for insert statements. If I add the executions column in the above statement, why do I see more rows processed then executions?

That's what I had based my original logic on. Number of executions of a particular statement divided by the elapsed_time to get how fast an insert occurs. What am I missing?

--
mohammed

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
--
http://www.freelists.org/webpage/oracle-l