Autoscaling Intro

Need of Solution

• Saving up costs by provisioning only required resources and not wasting too much of resources.
• Improved application availability.
• Efficient resource utilization.
• Elasticity: Capability of application to adapt unpredictable traffic patterns.
• Fault tolerance and recovery: Either recover from the failure or not even see the failure.
• Simplified management of applications.

Kubernetes Scaling

• Cluster Scaling:

  • Shifting the number of nodes inside the cluster.
  • Expands cluster’s total capacity (CPU, RAM, GPU, Disk, …).
  • Expands cluster availability.

• Pod Scaling:

  • Expanding or shrinking the pods or replicas numbers or resources using HPA or VPA.
  • Expands application availability and efficiency.

Note

Classification of kubernetes scaling strategies is important for maintaining: balanced, efficient and resilient environments.

Manual HPA

• It involves the editing of kubernetes deployment manifest replica option and applying the changes.
• This is very repetitive and error prone and can lead to disasters when wrong values are set.
• Scaling can changes the behaviour of the applications and their efficiency.

Manual VPA

• When each replica or pod needs more resources, they need to be upscaled vertically by adding more resources to all the instances
• this recreates all the instances and recreates them or replaces them with new instances and can cause some downtime and can break application availability