Project on inserts bench
Ruslan Zasukhin
sunshine at public.kherson.ua
Tue Nov 21 16:36:19 CST 2006
On 11/21/06 1:57 PM, "Ruslan Zasukhin" <sunshine at public.kherson.ua> wrote:
Okay David,
It seems I have found that INSERT with biding speed is broken in 2.5 b20
it not use pool of queries.
So we will fix it back and I will remake below bench.
>From your numbers old binding was 0.8 sec vs 7.7 of b20.
So below result must go to about 2.7 second ... Okay.
> Hi Guys,
>
> I have to see that this project was broken a little for last 2 buttons.
> VarChar field must be commented there. Attached is corrected project.
> If you want VarChar field also, it must be uncommented in few places.
>
> This project is interesting because it show
> FIVE (5) different ways to add records.
>
>
> ---------------
> So I have self take this old my project, which I use for profiling of engine.
>
> To use it you need do:
>
> click CREATE DB
> click Open DB
> Now click one of buttons in the second row.
> From left to right buttons do job faster.
>
>
> Now project insert 100K records. 8 field tables. Only numbers.
> I have this results:
>
> INSERT EXEC 38.00 sec
> InsertNoFields 29.37
> Insert Bind 26.13 sec
> Insert Add1 5.13
> Insert Add2 1.7 sec
>
> So we can see that SQL Binding improve speed about 25%.
> Still this looks not best because we can see grammatical
> Speedup if to use API way and Vtable, i.e. non-SQL way.
>
> The best difference is 38 / 1.7 = about 20 times.
>
>
> --------
> I.e. If you have some task on 44 minutes with simple INERTS,
> You can expect go down to 33 minutes with INSERT + BINDING,
> And down to 2 minutes with API way.
>
--
Best regards,
Ruslan Zasukhin
VP Engineering and New Technology
Paradigma Software, Inc
Valentina - Joining Worlds of Information
http://www.paradigmasoft.com
[I feel the need: the need for speed]
More information about the Valentina
mailing list