map performance

Stephen Woodbridge woodbri at SWOODBRIDGE.COM
Wed Dec 6 11:58:13 EST 2006


Boice,

You can take the post-generated mapfile and run that via the CGI with 
DEBUG TRUE to get stats on how long it takes mapserver to render each 
layer. This is the best place to start when you are trying to optimize 
stuff. You really need to understand what layers are taking your time 
and then look specifically at what you are doing with that layer. Make 
changes and retest, until you are happy with the results. Then move on 
to the next worst performing layer. This way issues and tips can be 
specific to your cases. I think we have given you all the "generic" 
advice already.

-Steve W

boice tomlin wrote:
> 
> Hi Ed,
> 
> Yes.  I'm aware of where PHP fits in.  I was just trying to make it 
> clear that I've not overlooked the PHP part of the process and don't 
> have any bottlenecks there.
> 
> Attached are both of the map files.  One is pre processing and post.  It 
> looks mostly as I expected.  Your input would be greatly appreciated.
> 
> I'm open to different approaches.  I hear people mention creating their 
> own shape files or modifying their shape files directly.  How could I 
> benefit from that?  Also I can create dbf files dynamically.  I'm 
> guessing I can do a join with those directly from the map file.  I don't 
> know if there would be any advantage from that.
> 
> thanks,
> 
> boice
> 
> 
> 
> =
> ------------------------------------------------------------------------
> 
> 
> 
> ------------------------------------------------------------------------
> 
> 
> 
> 
> 
> 
> 
> On Dec 6, 2006, at 7:54 AM, Ed McNierney wrote:
> 
>> Boice -
>>
>> I want to make sure you realize that your "PHP work" is really doing
>> nothing more than creating the data structures to be used in rendering
>> your map.  It's not reading any shapefiles, drawing output images, etc.
>> - it's just looping and creating instructions for the "draw" method to
>> follow.  The fact that it happens "really fast" is both unsurprising and
>> unimportant.
>>
>> It's as if you were running a program to generate instructions for
>> yourself to follow; generating a list that says "scratch your ear" 500
>> times is fast to generate and (relatively) fast to perform, but
>> generating a list that says "fly to Patagonia" 500 times is equally fast
>> to generate but enormously slower to perform.
>>
>> Your fast code is simply generating the list.  The map-drawing is doing
>> the work of following the instructions, so Steve's suggestion is still a
>> good one.  If you can use the $map->save() method and post the results,
>> then we can see what the instructions are.
>>
>>      - Ed
>>
>> Ed McNierney
>> President and Chief Mapmaker
>> TopoZone.com / Maps a la carte, Inc.
>> 73 Princeton Street, Suite 305
>> North Chelmsford, MA  01863
>> Phone: +1 (978) 251-4242
>> Fax: +1 (978) 251-1396
>> ed at topozone.com <mailto:ed at topozone.com>
>>
>> -----Original Message-----
>> From: UMN MapServer Users List [mailto:MAPSERVER-USERS at LISTS.UMN.EDU] On
>> Behalf Of Boice Tomlin
>> Sent: Tuesday, December 05, 2006 10:52 PM
>> To: MAPSERVER-USERS at LISTS.UMN.EDU <mailto:MAPSERVER-USERS at LISTS.UMN.EDU>
>> Subject: Re: [UMN_MAPSERVER-USERS] map performance
>>
>> Thanks Steve,
>>
>> I will save the map file and take a look.  great tip.
>>
>> In my case I have a few layers which are always on because I know the
>> get displayed on every request.  For instance I have on layer called
>> county. 
>> Every contry rendered on the map is from that single layer.
>>
>> What I do is grab all county data that are complete(a user has complete
>> data for the chosen layer in that geography) - green - and form a long
>> regular expression with all thouse counties.  Then I form a class
>> dynamically and set the color to green.
>>
>> Then I do the same thing for in-progress data, form a regular expression
>> from several hundred results, create a class from that same county layer
>> and set the color to orange.
>>
>> I'm not sure Steve if I'm hearing you right.  You may be referring to a
>> completely different method that I'm not getting.  The thing is that the
>> PHP work described above happens fast, really fast.
>>
>> The amount of time the maps takes to render is proportionate to the
>> amount of counties, cities, etc. in my regular expression.
>>
>> does that make my situation clearer.
>>
>> thank you for your help,
>>
>> boice
>>
>>
>>
>>> Might be interesting to see what the resulting mapfile (after your 
>>> dynamic work) looks like (use $map->save(...)).
>>>
>>> We'd really need to know more about the data, how you're doing 
>>> classifications and such to comment more. There may be lots of ways to
>>
>>> make things go faster. From the looks of it for each geography (city, 
>>> county, tribe, state, nation) you have a bunch of variables that 
>>> indicate if a data theme is available. One could organize that data 
>>> like so (for example state level data):
>>>
>>> State  Hydrography  Watersheds ...
>>> MN     0                    1
>>> WI      3                    2
>>> IA       3                    0
>>> FL       1                    3
>>>
>>> Where 0 means no responce, 1 complete and so on. So in that case your 
>>> class definitions would always be the same you'd be simply changing 
>>> the variable you're mapping on (e.g. CLASSITEM), and you wouldn't need
>>
>>> dynamic classes (or even MapScript for that matter).  Just thinking 
>>> out loud...
>>>
>>> Steve
>>>
>>>>>> boice tomlin <boice at RUNSKIP.COM <mailto:boice at RUNSKIP.COM>> 
>>>>>> 12/4/2006 1:24:58 PM >>>
>>>
>>> Hello users,
>>>
>>> The map located here;
>>>
>>> http://gisinventory.net/status_maps.html
>>>
>>> takes a while to load.  At least at the national view where there is a
>>
>>> lot of area to render.
>>>
>>> I am looking for alternative ways to generate the map that will 
>>> significantly improve performance.
>>>
>>> Currently I am using php and looping through data and turning on 
>>> layers as I go.  The PHP part is lightning fast.  But after I get the 
>>> map ready mapserver takes several seconds to generate it.
>>>
>>> I'm curious about alternative ways to handle this problem and wonder 
>>> if anyone had comments on any of them.
>>>
>>> 1) modifying the shape files in some way so the layer information is 
>>> in those files so that all mapserver has to do is load those files and
>>
>>> not depend on the map files.
>>> 2) using a db such as postgres with postgis so that layer information 
>>> is available all in one compact source.
>>>
>>> Right now I have to generate a bunch of dynamic classes in PHP using 
>>> the general method below.
>>>
>>> $lyr = $this->ramona_map->getLayerByName("state_yes");
>>>                 $cla = $lyr->getClass(0);
>>>                 $cla->setExpression("/".$expression."/");
>>>                 $lyr->set("status", MS_ON);
>>>
>>> I have to do this several hundred times to represent all of the data.
>>> The time it takes to generate the map seem proportionate to the amount
>>
>>> of layers I make visible.  And again this is on the mapserver side and
>>
>>> not PHP.  PHP does its part of the operation in thousandths of a 
>>> second.
>>>
>>> anyone's thoughts are greatly appreciated.
>>>
>>> -boice tomlin
>>>
>>>
>>>
>>> ////////////////////////////
>>> Run Skip
>>> http://runskip.com/
>>>
>>>
>>> boice tomlin
>>>
>>>
>>> boice at runskip.com <mailto:boice at runskip.com>
>>>
>>>
>>> 503-528-6204
>>>
>>
>>
>>
>> ////////////////////////////
>> Run Skip
>>
>> boice tomlin
>>
>> boice at runskip.com <mailto:boice at runskip.com>
>>
>> 503-528-6204
> 
> 
> 
> 
> ////////////////////////////
> Run Skip
> http://runskip.com/
> 
> boice tomlin
> 
> boice at runskip.com <mailto:boice at runskip.com>
> 
> 503-528-6204
> 
> 
> =



More information about the mapserver-users mailing list