Jody Garnett ha scritto:
> Andrea Aime wrote:
>> Emily Gouge ha scritto:
>>  
>>> Andrea,
>>>
>>> I updated my geotools and deployed it locally and have been using it 
>>> in UDig.  It seems to be working just fine for me - I haven't run 
>>> into any issues.  I'm going to continue hammering shapefiles in uDig 
>>> over the next couple weeks so it'll get some more testing from me.  
>>> I'll let you know if any issues arise.
>>>
>>> I can't really comment on the speed; it doesn't seem a lot faster to 
>>> me but I don't have any good comparisons.
>>>     
>> Well, maybe you don't have 40 threads trying to read the same file in
>> parallel. 
> Emily is working on tiles; so she may end up having multiple threads 
> accessing the data store in parallel. I recall andrea that you did some 
> fancy stuff with Jobs to prevent many disk bound renderers executing at 
> once? Do you recall how that worked ...

Yeah, that's a little different. At the time I noticed uDig was 
seriously thrashing the poor disk trying to access n shapefiles in 
parallel (in the customer case, n > 20) and I introduced the concept
of "zones", each zone was forbidden to have more than n threads playing
in it, much alike the browser rule not to hit the same server with more
than two connections at a time. For the local disk zone I believe the
limit was two. Implementation is probably less sophisticated than that,
an probably just limiting disk access.
However, tiling is another kind of beast, it accesses the same file
and in similar areas. If you're zoomed in enough the threads rendering
the same file will probably end up playing in the file system cache,
the trouble is when you're trying to render a massive layer with 
multiple threads when zoomed out, each threads trying to read a 
different part of a 1GB data source. In that case thrashing is still
very much likely, you may want to try and avoid it.

>> And oh, btw, the patch affects direct access to the datastore (wfs 
>> requests), the shapefile-renderer was not affected in my tests neither.
>>   
> ShapefileRenderer also hits the ShapefileDataStore - would it not be 
> effected by your change?

Theoretically it should, but the renderer dodges most of the datastore
code, like it has its own geometry handlers, and accesses stuff very low 
level (as in using the shape readers and writers directly). That is, I'm
not totally sure it's playing with locking or not, the rendering tests
did not seem to be affected by this scalability issue.

Cheers

-- 
Andrea Aime
OpenGeo - http://opengeo.org
Expert service straight from the developers.

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Geotools-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geotools-devel

Reply via email to