[Mapserver-users] Storing database connections in a CGI session

[Benjamin Wragg]

Just to answer Bart's question about ArcIMS and caching. ArcIMS 3 and
3.1 (and likely 4.0) defiantly cached shapefiles when a MapService was
setup. This actually happened when you told ArcIMS to make available an
AXL file (like a map file in mapserver). It would read through the data
sources in the AXL file and cache the shapefile data sources. 

Regarding SDE data sources, as far as I know it never cached from SDE
data sources. What they did do in version 4.0 (could also be in 3.1) is
they setup a new way of connecting to SDE. It was called a "direct
connect" and if my memory serves me correctly it basically took some of
the load from the server process which is assigned to handle all that
connections requests. ArcIMS some how communicated more directly with
the database itself. This apparently speed up the connection. This was
about the time I left the 'dark side' and joined the 'rebels' in there
quest to bring peace and harmony to the galaxy so my knowledge has some
gaps in it from version 4.0 on :).

Cheers,

Benjamin Wragg

-----Original Message-----
From: mapserver-users-admin at lists.gis.umn.edu
[mailto:mapserver-users-admin at lists.gis.umn.edu] On Behalf Of Bart van
den Eijnden
Sent: Wednesday, 10 December 2003 6:26 AM
To: Doug Cates; mapserver-users at lists.gis.umn.edu
Subject: Re: [Mapserver-users] Storing database connections in a CGI
session


Hi Doug,

are you absolutely convinced there is no caching involved with ArcIMS?

All the performance tests we did against ArcIMS showed a huge difference

between the maximum response time, and the average response time, which 
indicates caching in my opinion. This even was the case with shapefiles 
where no database connection is needed.

Best regards,
Bart

On 09 Dec 2003 11:22:22 -0800, Doug Cates <doug at moximedia.com> wrote:

> Richard, Bart,
>
> I'm also VERY interested in this, as I'm hoping to set up a MapServer 
> / SDE implementation early in the new year. Can you send me some 
> information on your testing too, Richard?
>
> As far as ArcIMS goes, I know for sure that it sets up multiple 
> connections to SDE for each ArcIMS service when the service is 
> created, and keeps them open until the ArcIMS service is shut down. I 
> don't believe that there is any caching taking place.
>
> Is there any chance that something can be built right into MapServer 
> to do something similar, opening a connection pool when the first SDE 
> map is requested, and re-using the connections if they are present?
>
> Doug
>
>
> On Tue, 2003-12-09 at 10:45, Bart van den Eijnden wrote:
>> Hi Richard,
>>
>> what is the absolute difference in performance (time) between 
>> Mapserver
>> en
>> ArcIMS? And can you tell some more about how you did the tests
(number 
>> of
>> concurrent users, which tool etc.)?
>>
>> I think the difference is not only in the database connection, but 
>> also
>> in
>> the caching of data. Although I do not know how ArcIMS works exactly
>> behind the scenes ofcourse.
>>
>> An alternative to CGI sessions is the fastcgi approach. Frank 
>> Warmerdam did some work on this a few months ago, but I don't know if

>> there any useable results.
>>
>> See also: http://mapserver.gis.umn.edu/cgi-bin/wiki.pl?FastCGI
>>
>> Best regards,
>> Bart
>>
>> On Tue, 9 Dec 2003 19:29:32 +0100, Richard Holzmeier 
>> <Richard.Holzmeier at epost.de> wrote:
>>
>> > Currently i am testing MapServers load performance against ArcIMS
>> using
>> > SDE
>> > and OracleSpatial database connections. ArcIMS displaces MapServer
>> with a
>> > performance rate up to 7:1, certainly due to the overhead of 
>> > creating
>> a
>> > new
>> > session with every request. I´m thinking about using CGI sessions 
>> > to store opened database sessions for a while using a library such 
>> > as
>> CGI::SESSION
>> > (http://search.cpan.org/~sherzodr/). Unfortunately i´m fairly new 
>> > with C/C++ and Perl and have no idea of the complexity or 
>> > possibility to
>> implement
>> > this.
>> >
>> > If anybody has, i will appreciate responses of any kind. Thanks in 
>> > advance, Richard
>> >
>> >
>> > _______________________________________________
>> > Mapserver-users mailing list Mapserver-users at lists.gis.umn.edu
>> > http://lists.gis.umn.edu/mailman/listinfo/mapserver-users
>> >
>>
>>
>
> _______________________________________________
> Mapserver-users mailing list Mapserver-users at lists.gis.umn.edu
> http://lists.gis.umn.edu/mailman/listinfo/mapserver-users
>



-- 
  
_______________________________________________
Mapserver-users mailing list
Mapserver-users at lists.gis.umn.edu
http://lists.gis.umn.edu/mailman/listinfo/mapserver-users