User Service Capacity Requests

Overview¶

Authorized users of SlideRule Earth now have the ability to request dedicated nodes on the public cluster for their private authenticated use. To access this capability, a user makes a user service capacity request to the SlideRule Provisioner which responds by spining up a dedicated auto-scaling group within the subnet of the public cluster using the node_capacity and ttl parameters provided by the user. The user then accesses this private auto-scaling group by supplying their account username in the x-sliderule-service header of subsequent processing requests to the public cluster.

Constraints¶

• Only an organizational member or affiliate has the ability to request user service capacity on the public cluster.

• The requested node_capacity and ttl are still bound by the user limits imposed by the SlideRule Provisioner for a given organizational role.

• Once a user service capacity request has been made, the auto-scaling group cannot be modified with additional capacity until it is destroyed and recreated.

Provisioner APIs¶

The following APIs are exposed by the SlideRule Provisioner for managing user service capacity requests. In all cases, the {username} provided must match the subject claim of the authorization token. When the cluster parameter of the request is “sliderule” then the Provisioner will automatically identify the request with the public cluster.

• /deploy/{username} - provision a user auto-scaling group within the provided cluster using the provided node_capacity and ttl

• /extend/{username} - extend the lifetime of the user auto-scaling group within the provided cluster using the provided ttl

• /destroy/{username} - tear down the user auto-scaling group within the provided cluster

• /status/{username} - respond with the status of the user auto-scaling group within the provided cluster

• /events/{username} - response with the cloud formation events associated with the user auto-scaling group within the provided cluster

Python Example¶

Here is an example Python code snippet that uses the SlideRule Python Client to provision user service capacity on the public cluster and then use it.

from sliderule import sliderule
session = sliderule.create_session(node_capacity=1, user_service=True)
sliderule.get_version(session=session)

Motivation¶

Previously, the only way for a user to have dedicated compute resources was to provisioner an entire private cluster. Private clusters are associated with a subdomain that routes via DNS to the load balancer of the cluster. In some instances, DNS record propagation was taking an extremely long time which rendered the private cluster unreachable. The user of auto-scaling groups inside the public cluster subnet alleviates the need for modifying DNS records and supports very fast and deterministic access to the compute nodes.

Private clusters are still available and useful for long running operations and operations accessed from within AWS infrastructure. Private clusters can also be leveraged for customized processing environments dedicated to a specific group.