Autoscaling

In computing, autoscaling is when systesm without manual intervention adjust capacity to meet the traffic demand by adding or removing resources.

Pod Based Autoscaling

There are two ways in which pod based autoscaling can be used

• Horizontal Pod Scaling (Add more pods)

• Vertical Pod Scaling (Increase CPU, memory capacity)

Horizontal Pod Scaling:

In this method we increase or decrease the number of pods in the node based on demand

Horizontal Pod Scaling

Vertical Pod Scaling:

In this method we increase or decrease the capacity/size of resources like CPU, Memory etc based on demand

Verticle Pod Scaling

Node Based Autoscaling:

• Known as ClusterAutoscalar or Karpenter

• We increase or decrease the number of nodes based on demand

Node Based AutoScaling

Scale V/S AutoScale

• Scale :

  • Update the amount of replicas in the state of deployment object

  • Perform Deploy

  kubectl scale --replicas=3 deploy/<app-name>

• Autoscale:

  • The autoscale command is used to create a HorizontalPodAutoScalar

  kubectl autoscale rc --min=1 --max --cpu-percent=80

KEDA - Kubernetes Event Driven AutoScaling

• Active MQ

• Apache Kafka

• AWS CloudWatch/ Kinesis/ SQS Queue

• Azure App insights/ Blog Storage/ Event Hubs/ Log analytic/ Monitor/ Pipeline

• New relic

• Prometheus

• Datadog