Perfomance has been noted to be a concern, especially in the current downsampling/writing techstack built on zarr-python_3 and dask. We should think about supporting other backends (tensorstore?), but this should really be driven by some sort of actual benchmarking. This could be inspired by the work of @dstansby over at this project.
In fact, maybe we could have this running on CI? Haven't done much work like this though.
cc @kevinyamauchi
Perfomance has been noted to be a concern, especially in the current downsampling/writing techstack built on zarr-python_3 and dask. We should think about supporting other backends (tensorstore?), but this should really be driven by some sort of actual benchmarking. This could be inspired by the work of @dstansby over at this project.
In fact, maybe we could have this running on CI? Haven't done much work like this though.
cc @kevinyamauchi