Is your feature request related to a problem? Please describe.
Currently, when ds-ml-service is deployed, both training and prediction services are just running flask. The recommended and minimal practice for being production ready is to make use of a WSGI server implementation, so to later (if required: caching, high request loads,...) being able to proxy it through a proper HTTP server via HTTP or socket connections.
Describe the solution you'd like
Make use of an already existing WSGI implementation (e.g.: uWSGI or Gunicorn) and spinup flask through it.
Describe alternatives you've considered
Also loading an HTTP server on top (e.g.: Nginx) --> out of scope since there is not expected high loads nor cache required (only API) <-- not required
Additional context
The implementation should be as transparent as possible to current ds-ml-service design.
Is your feature request related to a problem? Please describe. Currently, when ds-ml-service is deployed, both training and prediction services are just running flask. The recommended and minimal practice for being production ready is to make use of a WSGI server implementation, so to later (if required: caching, high request loads,...) being able to proxy it through a proper HTTP server via HTTP or socket connections.
Describe the solution you'd like Make use of an already existing WSGI implementation (e.g.: uWSGI or Gunicorn) and spinup flask through it.
Describe alternatives you've considered Also loading an HTTP server on top (e.g.: Nginx) --> out of scope since there is not expected high loads nor cache required (only API) <-- not required
Additional context The implementation should be as transparent as possible to current ds-ml-service design.