[Tilecache] SLD Performance

[Christopher Schmidt]

On Tue, Feb 13, 2007 at 10:24:58AM +1100, bradleyspencer wrote:
> I have built a demo that uses SLDs to render WMS layers atop of GoogleMaps
> in an OL app.
> 
> See http://www.cubewerx.com.au/DemoDrapesOL.html
> <http://www.cubewerx.com.au/DemoDrapesOL.html>
> 
> This is as good as I can get it with buffer: 1 and 512 x 512 tile sizes.
> 
> However, I would like to improve performance for the initial fetching of the
> remote tiles if possible. I understand that this could quite possibly be
> server-side issues associated with rendering on the fly and maybe this is as
> good as I can expect. But I wondered what could be done at the TileCache
> population level within OL?
> 
> Does anyone know of any way to improve the fetch times?
> Is OL serialising the WMS tile requests?
> If so is it possible to configure OL to launch off all the WMS tile requests
> in parallel to overlap the seeks times?

Browsers will only request 2 (or sometimes a small number more) tiles
from a single host at a single time. The answer to this is to use
multiple hosts. With the most recent version of OpenLayers, you can do
this by setting your WMS URL to instead be a set of multiple URLs, a la:

[http://tile1.cubeserv.org, http://tile2.cubeserv.org]

You can see this with Google, loading tiles from mt1.google.com,
mt2.google.com, mt3.google.com, etc.

If you have multiple layers you want to load simultaneously, you should, 
or at least can, set up different DNS aliases for each set of layers.

These can all point to the server: the limitation is on hostname, not
IP.

Regards,
-- 
Christopher Schmidt
MetaCarta