Manifests

Autoscaling

Autoscaling is a crucial feature for managing the scalability and efficiency of applications running in dynamic environments. For NitroEnclaveDeployments, autoscaling is implemented through a sophisticated controller that interacts directly with Prometheus to fetch real-time metrics. This custom autoscaler utilises a defined set of metrics, such as request rates or computational loads, to make informed decisions.

Horizontal NitroEnclave Scaling

A HorizontalNitroEnclaveAutoscaler automatically updates a NitroEnclaveDeployment to scale it to match demand.

Horizontal scaling means that the response to increased load is to add more Nitro enclaves. If the load decreases and the number of NitroEnclaves exceeds the configured minimum, the HorizontalNitroEnclaveAutoscaler instructs the NitroEnclaveDeployment to scale back down.

The HorizontalNitroEnclaveAutoscaler is implemented as a Kubernetes API resource and a controller. The definition of the resource determines the behaviour of the controller. The HorizontalNitroEnclaveAutoscaler, running as a controller pod, periodically adjusts the desired scale of the NitroEnclaveDeployment to match observed metrics such as average CPU utilisation, average memory utilisation, or any other custom metric you specify.

The Horizontal Nitro Enclave autoscaling is implemented as an intermittent control loop. Once during each period, the controller queries the resource utilisation against the specified metrics in each HorizontalNitroEnclaveAutoscaler definition. These metrics are fetched from Prometheus, where all the enclaves are aggregated.

Example: Starting a Horizontal Nitro Enclave Autoscaler

The HorizontalNitroEnclaveAutoscaler is a custom resource that automatically scales the number of Nitro Enclave Deployments based on observed metrics.

Here’s an example of a HorizontalNitroEnclaveAutoscaler for a FastAPI Nitro Enclave Deployment.

# hello-fastapi-deployment-hnea.yaml

apiVersion: autoscaling.example.com/v1
kind: HorizontalNitroEnclaveAutoscaler
metadata:
  name: hello-fastapi-deployment-hnea
  namespace: default
spec:
  scaleTargetRef:
    apiVersion: k8s.oblv.com/v1alpha1
    kind: NitroEnclaveDeployment
    name: hello-fastapi-deployment
  minReplicas: 1
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 50
    - type: Resource
      resource:
        name: memory
        target:
          type: Utilization
          averageUtilization: 70

Fields

To better understand the fields of the example manifest, here's a brief description:

General Information

Field	Description	Value
apiVersion	Specifies the API version for the custom autoscaler.	autoscaling.example.com/v1
kind	Specifies the kind of Kubernetes resource.	HorizontalNitroEnclaveAutoscaler

Metadata

Field	Description	Value
name	Specifies the name of the HorizontalNitroEnclaveAutoscaler.	hello-fastapi-deployment-hnea
namespace	Specifies the namespace where the autoscaler will be created.	default

Spec Configuration

Field	Description
spec	Contains the detailed specification for the HorizontalNitroEnclaveAutoscaler.

Scale Target Reference

Field	Description	Value
apiVersion	API version of the target.	k8s.oblv.com/v1alpha1
kind	Kind of target resource.	NitroEnclaveDeployment
name	Name of the target resource.	hello-fastapi-deployment

Scaling Parameters

Field	Description	Value
minReplicas	Minimum number of replicas.	1
maxReplicas	Maximum number of replicas.	10

Metrics for Scaling

Field	Description	Value
type	Type of metric used for scaling.	Resource
resource	Specifies the resource to be measured.	cpu
target type	Specifies the scaling target type.	Utilization
averageUtilization	Target average CPU utilisation percentage.	50

Usage

To create a HorizontalNitroEnclaveAutoscaler, you can use the kubectl apply command:

kubectl apply -f hello-fastapi-deployment-hnea.yaml

To view the current status of a HorizontalNitroEnclaveAutoscaler, you can use the kubectl get command:

kubectl get hnea hello-fastapi-deployment-hnea

To delete a HorizontalNitroEnclaveAutoscaler, you can use the kubectl delete command:

kubectl delete hnea hello-fastapi-deployment-hnea