List view
We could use ray as a cluster manager. Request spark executors from ray java API directly. In this way, we could request executor more flexible and don't need to startup a standalone cluster. Benefits: 1. more flexible to submit spark applications. 2. reduce the time for startup spark standalone cluster. Possible deficiencies: 1. cross language call is not perfect in ray.
No due date•3/3 issues closed