Combiner storage space issue/ model retention policies on the combiner

[saadiabadi]

Is your feature request related to a problem? Please describe. We are facing some problems with the FEDn in the combiner level where my combiners are running out of space because of the received model from all clients; those local models are saved in the Tmp file, which filled the combiner storage very fast, especially for large models and long-running experiment. for this reason, the combiner will be broken and the fedn network will not be available anymore.

Describe the solution you'd like

What I would suggest is to have model retention policies on the combiner level to remove all of these saved models after combining them and the global partial model is extracted

How would the solution positively affect the functionality? This will improve the combiner efficiency and reduce the required resources for each combiner, also it will keep the fedn network functioning correctly for long- experiments and large models without any further storage space problem.

[ahellander]

@morganekmefjord I think this one would benefit from your input, before we implement any solution.

[ahellander]

There is a solution now in develop branch.

[ahellander]

In 0.5