Amazon EKS — Scaling

EKS scaling operates at two levels: pod scaling (how many replicas) and node scaling (how many nodes). Getting both right means your workloads handle traffic spikes while keeping costs down.

---

Topics Covered

#	Topic	Description
1	Cluster Autoscaler	ASG-based node scaling
2	Karpenter	Next-gen node provisioning
3	HPA & VPA	Pod-level horizontal and vertical scaling
4	KEDA	Event-driven autoscaling
5	Right-Sizing Workloads	Resource optimization
6	Hands-On Lab	Configure HPA + Karpenter
7	Practice Questions	Test your scaling knowledge

---

Scaling Layers

Layer	Tool	What It Scales	Trigger
Pod horizontal	HPA	Pod replicas	CPU, memory, custom metrics
Pod vertical	VPA	Pod resource requests	Historical usage
Node	Cluster Autoscaler	ASG desired count	Unschedulable pods
Node	Karpenter	EC2 instances directly	Unschedulable pods
Event-driven	KEDA	Pod replicas	SQS depth, CloudWatch metrics, etc.

---

Scaling Flow

1. Traffic increases → pods hit CPU/memory thresholds
2. HPA adds more pod replicas
3. New pods are Pending (no node capacity)
4. Karpenter (or Cluster Autoscaler) provisions new nodes
5. Scheduler places pods on new nodes
6. Traffic decreases → HPA removes replicas → Karpenter consolidates nodes

---

Start Learning →