performance with different data formats

[Howard Butler]

At 7:58 AM -0500 8/23/06, Mark Adams wrote:
>Howard, thanks for the detailed response. I've added further comments
>and questions inline:
>
>>Have you looked at implementing FastCGI
>
>I forgot to mention that we had implemented fastcgi on the mapserver
>server. However, we haven't noticed much or any performance improvement
>from this. Probably this means we're not doing something right, but
>there is little of no documentation to guide us. We can tell that PHP is
>running with fastcgi because multiple instances of PHP get created etc -
>but how can you know if mapserver is going through fastcgi and if the
>connections are being pooled? (I'm assuming, based on the performance we
>have, that connections are not being pooled in our case.)

Ensure that 'mapserv -v' shows SUPPORTS=FASTCGI.  Set your 
FastCgiConfig in your apache configuration to be something similar to 
step 3 of the FastCGI howto.  And the most important bit is ensure 
that 'PROCESSING="CLOSE_CONNECTION=DEFER" is set on the MAP, or else 
MapServer will not do *any* connection pooling (FastCGI or not).

>
>>I think the MapServer CGI wizards on this list would know how to have
>>MapServer send back XML for a query request.
>
>Bart's email suggested using the WFS interface and filter encoding to
>get back an XML response. Unfortunately changing our application to
>systematically use WFS even when the data is local would be a big
>programming effort for us at this time. Also, it seems a bit of a hack
>to have to do this?

I'm not an expert with this, but I also think you might be able to do 
something with MapServer's templating engine.  I'm hoping others with 
more expertise jump in here and explain an approach to take.  I'll 
refrain from interjecting my personal opinion on whether or not WFS 
is a hack or not ;)


>Based on this response, it seems like our best bet at this point is to
>concentrate on getting more out of fastcgi. Are there any documentation
>resources we could access for this?

<http://mapserver.gis.umn.edu/docs/howto/fastcgi> is the current doc. 
You also might try to catch me or Frank on IRC for quick response to 
specific questions.  If anyone else with FastCGI experience knows of 
information that needs to be added to the document, let Frank, Jeff 
McKenna, or I know and we'll make sure to add it.

Howard