How metashape divides tasks among network nodes?

[aniket@aus.co.in]

I have a few questions on how the task distribution happens between nodes:
1. for different steps starting with Matching photos -> Aligning cameras -> Building dense cloud -> Building Mesh -> DEM -> ortho how these are bifurcated to smaller task?


If you have any article on this works under the hood because we need to understand how resource allocation works for different task.


While running workloads over network processing I've noticed that some of the nodes complete their tasks and sit idle. We can't move forward in the processing because we need to wait for that step to complete.

2. if the capabilities of different nodes are different then will metashape take that into account and give bigger tasks to bigger nodes and smaller tasks to smaller nodes?

Example:
Node 1
16 core CPU
8 core GPU
16 GB ram

Node 2
4 core CPU
1 core GPU
8 GB ram

will both the node be considered to be at par for running the small tasks provided or bigger tasks will be assigned to Node 1

[Alexey Pasumansky]

Hello aniket,

Metashape is not considering the hardware properties of the nodes for the task distribution purposes. You can set Priority for the more powerful nodes to higher values, in case you want to have them considered on the first place for selection from multiple available nodes.

If some of the processing nodes do not have GPU available, it is recommended to switch the Capability of such nodes to "CPU" option, thus they wouldn't be used for GPU-supported tasks. But you should have at least one node with the powerful GPU, of course, otherwise it wouldn't be possible to proceed, if there are no nodes with Any (default) of GPU capability.

[aniket@aus.co.in]

Thanks alexey for the explanation.


Please refer to the first question

How tasks are are distributed for different steps

for example in match photos how photos are selected that need to be matched together.
similarly for align photos how groups are made for distribution and so on for other steps.