[mapguide-users] Two performance questions

[Zac Spitzer]

there's a really simple answer here, all you need to do is TRYt out
for your self when it comes to
unmanaged versus managed... the next version of mapguide caches the
information about source
files which make the difference completely irrelevant based on some
testing Tom did recently.

Performance tuning is a black art, but i have to stress the concept of
testing out the concepts
yourself... it's all about conceptual understanding, pragmatism,
analysis and cold hard stats

too often i see people doing heaps of work in this area, based on the
wrong assumptions

any layers filters or expression based theming with big datasets will
direly affect performance

i've done a heap of titling over the years with mapguide, the limits
are more server based than
anything of the tiling client side

keeping the darn server up and running whilst tiling is the hardest
thing, out of memory problems
and apache occasionally exiting are often problems i see

with a real browser (IE not IE) like firefox you can tweak the number
of connections per host quite easily

the recent comment about ACE having an issue with 62 thread maximum
might play a role here

personally I always wonder why mapguide never seems to scale up very
well on windoz,
ie consistently under load utilising all 8 cores of a machine at 100%

"Removing unnecessary attributes from the SDFs" is completely
irrelevant here IMHO..
sdf performance rocks

apache modules aint isn't worth touching either, the limits are
nothing to do with the HTTP side of things

webserver tuning in this area only applies to serving out cooked tiles
to heaps of clients, if you consider
than 99% ish of the load and time to render a tile in withing the
mgserver process, it becomes clear that
attempting to tune 1% of  the load is pointless as the maxium gain is
always going to be less than
a 1% gain

putting a new paint job on a old car won't make it run faster

once you have your tiles made, I would recommend taking mapguide out
of the picture
and serving your tiles directly via a webserver, nginx is great...
cache headers are even better!

I'm about to launch a project I have been working for a few years, for
that project
there is no mapguide server used in production, it's all hosted via S3
and Cloudfront

hope some of this helps

z


On Sun, Feb 22, 2009 at 1:39 PM, Jason Birch <Jason.Birch at nanaimo.ca> wrote:
> What symptoms are you seeing with MgCooker?  I wonder if you're running into the same issue that I did.  I wanted to test with a single zoom level, so I initially unchecked the base layer group to unselect everything, and then selected the individual zoom level I was interested in.  When doing this, the cooker failed to run because the top-level checkbox didn't automatically get re-selected.
>
> Note, you'd have to test to see if this tool is any faster than the JS tile seeder.  Each instance of MgCooker seems to use a single HTTP connection, and makes requests via the MapAgent.
>
> Ideally, the tile service would have a method like CreateTile that generates the tile without returning the image to the client.  The success/failure of the tile creation could be indicated by the HTTP status code, with detailed exception information returned via XML/JSON content.
>
> I also wonder if tile seeding could be done with a native connection, to avoid the HTTP and webserver overhead.
>
> I'd definitely be interested to hear the conclusions of your tests.  It would be great to summarize all of the information from this thread on the MGOS wiki and/or the MapGuide Best Practices Wiki.
>
> Jason
>
> ________________________________
>
> From: gingerbbm
> Sent: Sat 2009-02-21 4:39 AM
> To: mapguide-users at lists.osgeo.org
> Subject: RE: [mapguide-users] Two performance questions
>
>
>
> Thanks for all your thoughts. Removing unnecessary attributes from the SDFs
> is something we can definitely do.
>
> My set-up is MGE2009 on Windows under IIS6 and my questions are really about
> how best to configure MapGuide for this environment. I'm running through
> various tests but thought it best to canvas opinion amongst people on the
> list who've been through these things before.
>
> I've had a quick look at Kenneth's cooker but it doesn't seem to be working
> right. Need to find more time for this because if there's a clear winner
> between that and Zac's JS seeder I want to use the quickest, obviously.
>
>
>
> _______________________________________________
> mapguide-users mailing list
> mapguide-users at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/mapguide-users
>
>



-- 
Zac Spitzer -
http://zacster.blogspot.com
+61 405 847 168