Re: V3 protocol, batch statements and binary transfer

Andrea,

Feel free to chip in, if you can help with the V3 implementation your
patches would be greatly appreciated.

Dave
On Tue, 2004-03-30 at 03:44, Andrea Aime wrote:
> Hi jdbc driver hackers,
> my name's Andrea and I'm hitting some serious performance problem with the driver.
> At present I'm working with the Postgis GIS extension and I'm hitting very low
> performance during mass data insertion due to the driver limitations. Basically,
> I want to turn a 20 MB shapefile into a postgres table, but it has to be an import
> function on the client side (windows pc) so I can't just go to the command line and
> issue a copy. But that's just an example, in general I need to perform mass insert
> or updates in a transactional environment from a client, usually a Windows PC.
>
> As far as I can tell the low performance level is due to:
> a) lack of true support of batch statements as introduced by the V3 protocol, that
>     makes the network latency bite me very badly while I'm inserting that 100000 rows
> b) use of the text mode instead of the binary one, more than doubling the size of data
>     that are really transfered over the wire
>
> That makes the insertion of the above file take more than 2 minutes on a 100MB ethernet
> (oh, I have to pass thru 3 switches, so the latency is not that good). A
> reasonable transfer time for that amount of data should be less than 30 seconds IMHO.
>
> I'm wondering, why do you use the text mode instead of the more efficient binary one?
> Secondly, reading the e-mails on the archive it appears that you are short of time
> for implementing the V3 protocol. Can I help somehow?
>
> Best regards
> Andrea Aime
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 6: Have you searched our list archives?
>
>                http://archives.postgresql.org
>
--
Dave Cramer
519 939 0336
ICQ # 14675561