Posted by Asarum on 06/13/06 21:58

On Tue, 13 Jun 2006 23:37:37 +0200, Cray wrote:

> Hello,
>
>
>
> On the database we have stored procedure that loads couple thousands
> rows with information, it has to be done sequentially because during
> this load various variables and values has to be changed and modified.
> We have about 10000 rows to process every day - usually it takes about 1
> hour.
>
> Some time ago I have noticed a big decrease in performance of this
> process (it takes about 10 to 20 more time to process 10K), profiler
> showed me that duration of each read from some of the tables is huge and
> this is the reason why this process slows down. I have also a
> maintenance plan that updates statistics. When I run this job during
> processing usually everything goes back to normal and reads are very
> fast, performance is ok.
>
>
>
> Could someone explain why this is happening or give me a hint where I
> could find the explanation/answer.
>
>
>
> Greetings Cray
> ...........:::::::: [ @ ] ::::::::...........
> Cray at o2.pl

I would take a closer look to indexes - if you make inserts into a table
and if it grows substancially, then if it necessary to scan the whole
table to insert a single row, then performance can significantly drop down.

[Back to original message]