search

neracoos-open / NFWF

Code and issues for the NERACOOS NFWF project
MIT License
2 stars 1 forks source link

sci-wms performance #14

Open rsignell-usgs opened 7 years ago

rsignell-usgs commented 7 years ago

This request for a single tile from the SMAST NECOFS model takes about 2 minutes (!), when the ncWMS service is running on whoi.edu/gamone and accessing data via opendap from smast.umassd.edu: http://gamone.whoi.edu/wms/datasets/necofs_massbay?service=WMS&request=GetMap&version=1.1.1&layers=u,v&styles=vectors_cubehelix&format=image/png&transparent=true&height=256&width=256&colorscalerange=0.0,1.0&numcontours=20&logscale=false&srs=EPSG:3857&vectorscale=20&bbox=-7856503.515263555,5087648.602661332,-7846719.575643052,5097432.5422818335 to produce this tile:

necofs_massbay

kwilcox commented 7 years ago

Are scalar layers quicker?

kwilcox commented 7 years ago

They are: http://gamone.whoi.edu/wms/datasets/necofs_massbay?service=WMS&request=GetMap&version=1.1.1&layers=temp&styles=&format=image/png&transparent=true&height=256&width=256&colorscalerange=0.0,1.0&numcontours=20&logscale=false&srs=EPSG:3857&vectorscale=20&bbox=-7856503.515263555,5087648.602661332,-7846719.575643052,5097432.5422818335

rsignell-usgs commented 7 years ago

Wow, that's like 1 second or less, right? Why the huge difference?

kwilcox commented 7 years ago

Filledcontours does not draw every cell... the vector request is actually drawing every cell's vectors that is in view. It's also processing 2x the data, but I'm assuming the slowdown is the matplotlib drawing. A small zoomed in area should be quick.