Jobs

A unit of work in Anyscale

Jobs in Anyscale represent a unit of work. They can contain many tasks, actors, and all the code you want.

Running a Job in Anyscale

In open-source Ray, you can use whatever script you are running in Python as a driver to kick off work inside your Ray cluster. Your Ray cluster might be local (for development) or in a cloud (for larger workloads).

In Anyscale, things are no different. You won't need to make any code changes for things to "just work"!

Running a Job in Anyscale

Changing from your open-source Ray cluster to Anyscale's managed clusters is as easy as:

RAY_ADDRESS=111.222.101.202 python your_script.py # Open-source Ray
# change to...
RAY_ADDRESS=anyscale://my_cluster python your_script.py # Anyscale

You can also choose to set the cluster URL in your Python script:

ray.init(address="anyscale://my_cluster")

Both have the same effect.

All of the remote invocations in your script (any tasks, actors or Ray-based libraries) will be run on the cluster you specified.