Use Spot Instances to Cut Batch Workload Costs by Up to 70%

Spot Instances use spare EC2 capacity auctioned at discounts of up to 90% versus On-Demand pricing. AWS can reclaim them with a two-minute warning, making them ideal for workloads that can tolerate interruption.

When to Use Spot

• Data processing and ETL pipelines
• Machine learning training jobs
• CI/CD build agents
• Rendering and media transcoding
• Load testing environments

Mixed Instance Strategy

Use a mixed instance policy in your Auto Scaling Group — combine Spot with a small On-Demand base. Specify multiple instance families (m5, m5a, m5n) in the same capacity pool to reduce interruption frequency.

aws autoscaling create-auto-scaling-group \
  --mixed-instances-policy '{
    "InstancesDistribution": {
      "OnDemandBaseCapacity": 1,
      "OnDemandPercentageAboveBaseCapacity": 0,
      "SpotAllocationStrategy": "capacity-optimized"
    }
  }'

Handling Interruptions

Poll the EC2 IMDS endpoint for the termination notice:

http://169.254.169.254/latest/meta-data/spot/termination-time

On detection, checkpoint state and exit cleanly within the two-minute window. For Kubernetes, use the AWS Node Termination Handler to cordon and drain the node automatically.