PRISM is an interactive map-based dashboard that simplifies the integration of geospatial data on hazards, along with information on socioeconomic vulnerability
MIT License
45
stars
32
forks
source link
Exposure analysis request consumes near 4gb of memory on open data cube request #423
The current exposure analysis, which requests high resolution population raster data from WFP's data cube instance, highlights a problem we have with WCS requests used by PRISM. The population raster dataset from worldpop is 100 meter resolution, and we're requesting an uncompressed file with national coverage which, in the case of Mozambique, is 3.74.
As an example, a request for exposure to a tropical storm in Mozambique generates this request:
Of note, we'll have a STAC implementation in the near-term and should consider moving away from WCS requests and use STAC instead. Details on STAC are here and example of downloading a raster through a STAC request is here
The current exposure analysis, which requests high resolution population raster data from WFP's data cube instance, highlights a problem we have with WCS requests used by PRISM. The population raster dataset from worldpop is 100 meter resolution, and we're requesting an uncompressed file with national coverage which, in the case of Mozambique, is 3.74.
As an example, a request for exposure to a tropical storm in Mozambique generates this request:
In a test instance of our Data Cube, this fails with a 500 error:
And this is the error raised on the data cube:
We had a similar issue surface when we tried to develop a download feature for raster files. See this PR https://github.com/WFP-VAM/prism-frontend/pull/208 and @valpesendorfer 's comment
Of note, we'll have a STAC implementation in the near-term and should consider moving away from WCS requests and use STAC instead. Details on STAC are here and example of downloading a raster through a STAC request is here