[Benchmarking] Coordination

[Paul Ramsey]

On Fri, Jun 12, 2009 at 11:58 PM, Andrea Aime<aaime at opengeo.org> wrote:
> - what's the mix of stores for those layers? Shapefile, postgis,
>  SDE, Oracle?

Well, we could go whole hog. I think it mostly is a matter of doing
the appropriate installs and loading the data into documented
tables/layers.

I think we commit to Shape/PostGIS as usual and treat others on an "as
available" basis.

> - let's do again a large raster layer? Again MrSID, or shall we
>  use JPEG2000, or ECW? (last year quite a bit of people asked
>  me why I did not use ECW, the answer being that I was not
>  comfortable with licensing)

Yes, one large wavelet raster. We're going Down Under so maybe we
should do ECW this year?

> - what about an image mosaic with 100-1000 tiles?

Are the semantics similar enough for comparisons? More to the point,
do we have that quantity of data?

> - we could also take OSM data from a city and do a OSM like
>  styling of it? Maybe a bit simplified in order to keep
>  all the servers on the same line (e.g., if one server
>  cannot do styling X it may fall back on a simpler, faster
>  option, invalidating the comparison)

OSM has kewl points, but just styling TIGER is sufficient to get what
we want, which is a number.

This question is actually double-barelled:

- Multi-class road style? (Yes!)
- Multi-layer / multi-class full base map (maybe?)

It seems like a multi-layer test really should be added, since it's a
real-world use case that we don't currently exercise. But doing the
data/style set-up on it will be a pain.

> - using again the roads layer, what about a style dependent
>  on the zoom level test? Only highway scaled out, everything
>  deeper in?

No, I think scale dependence doesn't tell us much.

> - point layer with externally provided symbol, like an
>  SVG or a PNG icon?

Yes, that seems like a good idea.

> Oh, mosaic wise, there was quite a thread on OSGEO discuss
> about mosaics on the fs vs mosaic in a database. Well, GeoServer
> does happen to have both options now. I guess a comparison
> between the two could prove entertaining, even if it's a GS
> only show? (to be included right after the cross server mosaic test?)

I don't even know what "mosaic in the database" means, so I am a "no".

P

> Cheers
> Andrea
> _______________________________________________
> Benchmarking mailing list
> Benchmarking at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/benchmarking
>