[mapserver-dev] RFC97: Dynamically Creating High Zoom-Level Tiles

Stephen Woodbridge woodbri at swoodbridge.com
Tue Apr 16 07:51:31 PDT 2013


+1 great idea.

Duh! Told you I hadn't had enough coffee, got up and down flipped in my 
head.

Thanks,
   -Steve

On 4/16/2013 9:43 AM, thomas bonfort wrote:
> Steve,
> if max is 15, you would only resample for levels 16, 17 and up, i.e.
> always from a single tile using the common power-of-two grids (g, wgs84,
> ...) or at the worst from 4 tiles for the other ones.
>
> so z <= max-cached-level : usual behavior
> z > max-cached-level : upscale from tiles where z=max-cached-level
>
> clearer ?
>
> --
> thomas
>
>
> On 16 April 2013 15:36, Stephen Woodbridge <woodbri at swoodbridge.com
> <mailto:woodbri at swoodbridge.com>> wrote:
>
>     On 4/16/2013 2:43 AM, thomas bonfort wrote:
>
>         Devs,
>         Please find RFC97:
>
>         http://mapserver.org/__development/rfc/ms-rfc-97.html
>         <http://mapserver.org/development/rfc/ms-rfc-97.html>
>
>         A bit short for an RFC, but a bit too long for just a ticket...
>         This RFC
>         basically allows serving tiles from high zoom levels by
>         upscaling lower
>         zoom level ones at request time, thus avoiding to fill up the caches
>         with upscaled data.
>         Note that the RFC also provisions the usage of proxying these high
>         zoom-level tiles to the source WMS instead of reconstructing
>         them from
>         the lower level tiles (use-case: the WMS is fast enough when
>         zoomed-in,
>         but needs caching when far away). This isn't implemented yet,
>         but should
>         be a contribution by the OpenWebBlobe folks in some future.
>
>
>     Hi Thomas,
>
>     I have a couple of questions, probably not enough coffee to get my
>     brain turned on yet.
>
>     Your goal is to produce up sampled tiles from tiles at the “maximum
>     cached level” using the reassemble strategy.
>
>     So if that is level 15, then at 14 each tile would need to resample
>     4 tiles, and at 13 need to resample 16 tiles, and at 12 need to
>     resample 64 tiles, etc.
>
>     And these re-sampled tiles are not going to be stored in the cache?
>
>     I can understand that if this resampling is fast that that would
>     make sense, but it would seem to me that at some point the disk IO
>     would start to kill this not to mention the CPU load. Why would you
>     not want to store these tiles or take a hybrid approach and save
>     every 2nd or 4th level above the max as a compromise between storage
>     and performance?
>
>     Am I missing something obvious here?
>
>     -Steve W
>
>
>     _________________________________________________
>     mapserver-dev mailing list
>     mapserver-dev at lists.osgeo.org <mailto:mapserver-dev at lists.osgeo.org>
>     http://lists.osgeo.org/__mailman/listinfo/mapserver-dev
>     <http://lists.osgeo.org/mailman/listinfo/mapserver-dev>
>
>



More information about the mapserver-dev mailing list