WMS tile cache that works with time dimension

Question

Is there a good WMS tile cache that understands the WMS time dimension? And what configuration is necessary? Open source preferred.

I am serving some WMS layers with the time dimension, so the map requests from the client include dates. For example http://…?…request=GETMAP&…&TIME=2010-12-10 returns the map for 10 Dec 2010. In my undestanding the tile cache will need to treat the TIME parameter as an identifier for the maps, just like the layer name.

I'm aware of Geowebcache, Tilecache, and MapProxy but I can't find any reference to the time dimension in their documentation. The time dimension has existed since WMS 1.1.0 so I would hope it's been considered by someone? Our layers are served by ncWMS and MapServer, if that's relevant. We would like to use a tile cache to increase the speed for our most frequently requested content. Can anyone recommend a tile cache, and preferably point us to the relevant documentation too?

Answer 1

There was a mention of a TileCache patch for this on the OSGeo mailing lists:

http://www.osgeo.org/pipermail//tilecache/2009-June/001863.html

Unfortunately the links to the patches no longer work. Apparently the patch involved a few changes to the code base, and would allow different caches to be used based on any URL argument.

I have an old script (in need of some refactoring), that grabs a request and adds a few extra parameters via Python. You write something like this to change the cache location based on the time parameter:

https://bitbucket.org/geographika/mapserver-scripts/src/tip/tilecache/tilecache.py (and a blog post on setting it up in Windows)

If your data does not change every day, then it may make more sense to create caches for ranges of time (e.g. every 2 months), grab the TIME parameter, see which period it belongs to and then point it to the cache for that period. Otherwise you could end up with an enormous amount of files.

Finally, have you optimised your database and WMS server as much as possible? MapServer can serve data very quickly without a cache, and as long as you have your HTTP headers set correctly user's browsers will handle caching locally.