Data Privacy¶
Run:ai SaaS Cluster installation uses the Run:ai cloud as its control plane. The cluster sends information to the cloud for control as well as analytics. The document below is a run-down of the data that is being sent to the Run:ai cloud.
Note
If the data detailed below is not in line with your organization's policy, you can choose to install the Run:ai self-hosted version. The self-hosted installation includes the Run:ai control-plane and will not communicate with the cloud. The self-hosted installation has different pricing.
Data¶
Following is a list of platform data items that are sent to the Run:ai cloud.
Asset | Data Details |
---|---|
Workload Metrics | Workload names, CPU, GPU, and Memory metrics, parameters sent using the runai submit command |
Workload Assets | Workload Assets such as environments, compute resources and data resoruces |
Resource Credentials | Credentials to cluster resources are stored and encrypted using a SHA-512 algorithm. The encryption is tenant-specific |
Node Metrics | Node names and IPs, CPU, GPU, and Memory metrics |
Cluster Metrics | Cluster names, CPU, GPU, and Memory metrics |
Projects & Departments | Names, quota information |
Users | User Run:ai roles, emails and passwords (when single-sign on not used) |
Run:ai does not send deep-learning artifacts to the cloud. As such any Code, images, container logs, training data, models, checkpoints and the like, stay behind corporate firewalls.
See Also¶
The Run:ai privacy policy.