From: | "Merlin Moncure" <mmoncure(at)gmail(dot)com> |
---|---|
To: | femski <hypertree(at)yahoo(dot)com> |
Cc: | pgsql-performance(at)postgresql(dot)org |
Subject: | Re: Postgres batch write very slow - what to do |
Date: | 2007-03-16 12:48:22 |
Message-ID: | b42b73150703160548t683ee4eckc3b05275e000b02b@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On 3/15/07, femski <hypertree(at)yahoo(dot)com> wrote:
>
> I tried maxpages = 0 and full_page_writes=off and it seemed to be taking
> forever.
> All other tricks I have already tried.
>
> At this point I wondering if its a jdbc client side issue - I am using the
> latest 8.1.
> (as I said in an earlier post - I am using addBatch with batch size of 100).
> But just in case - I am missing something.
>
> If 17k record/sec is right around expected then I must say I am little
> disappointed from the "most advanced open source database".
Be careful...you are just testing one very specific thing and it its
extremely possible that the Oracle JDBC batch insert is more optimized
than PostgreSQL's. On my little pentium 4 workstation, by inserting
10 rows per insert:
insert values ([...]), ([...]), [8 more rows];
I got a 5x speedup in insert performance using this feature (which is
unfortunately new for 8.2). Oracle is most likely pulling similar
tricks inside the driver. PostgreSQL is much faster than you think...
merlin
From | Date | Subject | |
---|---|---|---|
Next Message | Merlin Moncure | 2007-03-16 13:30:40 | Re: Postgres batch write very slow - what to do |
Previous Message | Heikki Linnakangas | 2007-03-16 09:35:20 | Re: Determine dead tuples size |