A step back, even if we can use all the memory and all the CPU power on the host computer, they won't be enough anyway as we are dealing with massive dataset and heavy compute workloads.
My view on this regard is that we should think about a more scalable architecture to further separate UI with the compute part, to make it distributed, e.g. UI in browser, compute in ImageJ-server running remotely, and connecting them with RPC over flexible transport layer.
@ctrueden thanks for merging imagej/imagej-legacy#256 so quickly! In the travis build of
imagej-legacy, I see this now (presumably caused by the bump to
[ERROR] Cannot create plugin: class='io.scif.convert.FileToDatasetConverter', priority=1.0, enabled=true, pluginType=Converter
… whenever we look for any
Any idea what caused this?
DatasetIOServiceis missing in the test context.
required=false, and then return
supportswhen the service is missing?