[mapserver-users] Performance implications of OFFSITE usage?

[Ed McNierney]

I'm doing some testing with USGS DRG maps projected as UTM and
collar-clipped.  My initial test took the raw output files - one per DRG
- and used the OFFSITE parameter to cause the pixel value 0 (black) to
be transparent.  This allowed me to create a TILEINDEX file for the raw
DRGs and just let MapServer take care of the overlap.

I've also got a DRG mosaic/clipping tool that can take these raw,
collar-clipped DRGs, mosaic them together, then chop out rectangular (in
UTM space) tiles.  I can then load them into a new TILEINDEX shapefile
and avoid using the OFFSITE parameter because the tiles do not overlap.

Now that I'm halfway through this I'm doubting the wisdom of it.  My
motivation was that the use of the OFFSITE parameter was forcing
MapServer to examine the value of every pixel rather than just process
them in bulk.  Avoid this should be a performance win.  I'm now
wondering whether that's a bogus assumption; when rescaling the source
TIFFs to produce the output image, enough pixel-level processing must be
going on that the OFFSITE comparison could be a trivial impact.

If anyone's got sage advice, experience with the libtiff or mapraster
code that's doing this, or just random opinions <g>, I'd appreciate any
thoughts.  Thanks!

	- Ed

Ed McNierney
Chief Mapmaker
TopoZone.com
ed at topozone.com
(978) 251-4242