Before cutting 1.0.0, I think it would be better to slightly change context default settings.
Shared mode fits more for already well-tested functions to achieve better performance. At the start, it's better to use exclusive mode for development and testing because of #437.
Increasing parallelization level should be also explicitly done from the user side because it depends on spark cluster setup.
Before cutting 1.0.0, I think it would be better to slightly change context default settings.
Shared
mode fits more for already well-tested functions to achieve better performance. At the start, it's better to useexclusive
mode for development and testing because of #437. Increasing parallelization level should be also explicitly done from the user side because it depends on spark cluster setup.