I agree with the meta data but the data layout is not clear to me.
From our experience the real bottleneck handling (raw and processed) DAS data is I/O. I think that hdf5 is well suited for the job, but the layout of the data is very important. Please see a proposed object oriented structure here:
Where the DASFrame shall be limited to a number of samples in time, not number of channels (similar to miniseed). The methods guarantee that one can easily scan a data set without loading the data. Also one must consider memmaping the data.
I agree with the meta data but the data layout is not clear to me.
From our experience the real bottleneck handling (raw and processed) DAS data is I/O. I think that hdf5 is well suited for the job, but the layout of the data is very important. Please see a proposed object oriented structure here:
Where the
DASFrame
shall be limited to a number of samples in time, not number of channels (similar to miniseed). The methods guarantee that one can easily scan a data set without loading the data. Also one must consider memmaping the data.