[gdal-dev] clarifications anout gdal_wms, tilecache and gdaltindex

[G. Allegri]

Hello list
I'm facing for the first time the need of serving a big raster dataset
continuum though a Mapserver wms service. The data source, at the moment, is
an ArcIms service (this comes for various reasons).
I've considered two options to tile, cache and serve it:

 - use gdaltindex to use the tileindex functionality of Mapserver
 - use gdal_wms pointing to a TileCache wms-c service

In the latter way I've setup the xml configuration file to use the highest
resolution level available in the TC cache, and set the maximum serveble
bbox. Now, I suppose that the chain is the following:

 1 - mapserver passes the request extent and dimensions to the gdal driver
 2 - gdal_wms calculates the tiles needed to fullfill the request at the
level set in TileLevel
 3 - it turns the "tiled requests" to TC wich verify if they match the tiles
in the cache. If ok it give them back to gdal_wms
 4 - gdal_wms mosaic the tiles (and clip them to the extent of the MS
request?)
 5 - Mapserver returns the result

Is it right? If it is, where is the performance improvement respect to
tiling (with gdaltindex) a raster as big as the ensemble of the whole TC
tiles at it's highest resolution? I mean, I can have 20x20 tiles cache in TC
for my bbox at the maximum level chosen, let's say 5, that is 5120x5120
pixels (standard tiles). Now, if I generate a single image with the same
dimensions and then use gdaltindex to break it in tiles and serve them
through a tileindex, doesn't I obtain a similar result (in performance
terms) as using gdal_wms?

I'm sorry if I haven't been very clear, but I hope you have understood my
doubts.

Thanks a lot for any suggestion,
Giovanni
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/gdal-dev/attachments/20090917/2b782ecc/attachment.html