Kubernetes Cost Optimisation: Measure Before You Change

Start a cost review with workload measurements and reliability requirements. Moving to Kubernetes does not guarantee savings, and no fixed percentage reduction applies to every environment.

Review requests and limits

Compare resource requests with observed demand, including peaks. Kubernetes uses requests when scheduling workloads; CPU limits can throttle execution and memory limits can lead to termination. Change settings gradually and measure the effect.

Plan for interruption

Spot capacity can suit interruption-tolerant workloads when retries, checkpoints and spare capacity are designed in. Interruption notices are best effort; do not depend on always receiving the full warning period.

Match capacity to demand

Review idle nodes, unused resources and scaling behaviour. Check that savings do not simply move cost into slower responses, failed jobs or greater operational effort.

Calculate storage costs

S3 Intelligent-Tiering can help with changing access patterns, but object size, monitoring charges and access requirements affect the result. Compare the actual storage profile rather than promising a standard saving.

Report the outcome

Compare bills over equivalent periods and record traffic, latency and error rates alongside cost. Make the assumptions and tradeoffs visible.

Sources and further reading