[mapguide-internals] Raster provider problems...

[Frank Warmerdam]

Traian Stanev wrote:
> Yes I understand all that.
> 
> But mostly working is way better than none at all. That's all I was trying
> to say.
> 
> That, or every request for a raster -- even for formats that would work fine
> -- will have to pass through the same critical section in order to
> synchronize access. I wouldn't call that an enterprise class solution
> either.

Traian,

Frankly I don't see this as a major impediment.  Raster access is generally
io limited, so having two threads accessing rasters off the same physical
disk is likely to result in seek-thrashing as much as it is likely to help
performance.

So, in my not-so-humble-opinion, single threading the fdogdal provider is
not going to appreciably reduce throughput - assuming that other non-fdogdal
activity in mapguide can continue for other requests.

In some circumstances - for instance with raster data spread over multiple
physical disks, or for compressed formats where there is a lot of processor
overhead to decompress, multi-threading might help significantly on a multi
processor system.

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | President OSGeo, http://osgeo.org