However we also believe the above approach might change depending on how we scale the process. (It takes 12.5 hours to measure a ~140M pixel image, which is way too slow!) So we aren't sure what the sizing will be.
A temporary solution could be to dramatically over-provision storage for now. Balanced PDs in us-central1 cost $0.10 per GB per month. So, 500 GB would cost $50 for the month. If we "only" ran for 0.5 days, that would be ~$0.84. That's cheap enough to not worry about it while we scale up.
Today, the qupath code uses the disk for the following:
For large enough images, this quickly exceeds the 30GB default boot disk.
As we did in https://github.com/dchaley/deepcell-imaging/issues/305, we could figure out the space requirements then add the appropriate disk.
However we also believe the above approach might change depending on how we scale the process. (It takes 12.5 hours to measure a ~140M pixel image, which is way too slow!) So we aren't sure what the sizing will be.
A temporary solution could be to dramatically over-provision storage for now. Balanced PDs in us-central1 cost $0.10 per GB per month. So, 500 GB would cost $50 for the month. If we "only" ran for 0.5 days, that would be ~$0.84. That's cheap enough to not worry about it while we scale up.