Distribution + Scaling
In order to ensure our software is scalable and portable we have developed an advanced distribution and scaling strategy.
We have developed this strategy using the Ray library, an open-source framework to build and scale Python applications easily.
It is particularly well-suited for applications that require high performance on large clusters, allowing us to scale our software to handle large and complex inversions. Before running an inversion with Astrape you must start a Ray cluster. This can be done with the following command:
ray start --head
Once the Ray cluster has been started, the distribution and scaling of the inversion process will be handled automatically by Astrape in a way that is most efficient for the cluster.
Once a Ray cluster has been started the Ray dashboard can be accessed by connected to the correct IP address and port. This dashboard provides a real-time view of the cluster’s status and resource usage, as well as a log output from the inversion process, allowing you to monitor the inversion as it progresses.
Stopping Jobs
Once inversion jobs have been submitted to the Ray cluster, they can be stopped using the following command:
ray job stop <job_id>
This will stop the job with the given job ID.
Stopping the Ray Cluster
Once the Ray cluster has been started, it can be brought down using the following command:
ray stop
This will stop the Ray cluster and release all resources.