CloudBender users can now granually configure access to regional datastores, data connectors, and services on a per project basis.

You can now securely load secrets or other key material into a trainML Project for use in jobs. This avoids the insecure method of providing credentials or other sensitive information in the form of environment variables.

trainML now support AWS as a CloudBender cloud provider. Start proxiML Clean Rooms directly in your AWS regions today!

GCP CloudBender regions now support utilizing Google Persistent Disk block devices as regional datastores. trainML particularly recommends this datastore type to be used for Volumes that need the low latency performance of an attached block device.

trainML as removed the requirement for cloud provider CloudBender regions to use Central storage mode. You can now create CloudBender regions comprised of only compute nodes.

CloudBender workloads can now enable communication with regional network-based resources using Regional Data Connectors. Currently, only IP/Port level white-list connectors are supported, but more are in development.

CloudBender Reservations has been rebuilt and renamed to Services. This restructure adds support for publicly addressable HTTPS service attachments to CloudBender workloads and well as the ability to issue publicly trusted certificates to private services (previously port reservations).

trainML jobs can now be configured to securely communicate with private Regional Services deployed in the same CloudBender region.

trainML now supports the ability to attach writable volumes to jobs to serve as persistent storage to Endpoints.

Physical CloudBender nodes are now substantially easier to provision. Instead of requesting a USB key, a CloudBender provisioning ISO can be downloaded directly from trainML website.