What it adds up to.
WHAT DRIVES COST
Compute for nodes Storage, including volumes and snapshots Load balancers Data transfer Registry storage Control plane charges, on managed clusters
WHAT IS FREQUENTLY WASTED
Nodes at low utilisation Over-generous resource requests Load balancers per service Volumes retained after workloads are removed Old images never cleaned
WHY REQUESTS DRIVE COST
The scheduler places by request, so generous requests mean more nodes than needed.
WHAT TO DO ABOUT IT
Measure actual usage and set requests to reflect it.
WHAT TO CONSOLIDATE
External traffic behind one ingress rather than many load balancers.
WHAT TO AUTOMATE
Cleanup of unused volumes and images.
WHAT TO MONITOR
Cost per environment Utilisation per node Idle workloads
WHAT NON-PRODUCTION ENVIRONMENTS COST
More than expected, because they run continuously and are used occasionally.
WHAT TO DO
Scale them down outside working hours.
WHAT TO CONSIDER LOCALLY
Pricing in foreign currency Transfer charges on pulling large images
WHAT REDUCES THE SECOND
Smaller images, and a local mirror.
WHAT TO REVIEW QUARTERLY
Whether the platform's cost is justified by what it provides.
WHAT TO BE HONEST ABOUT
Whether simpler infrastructure would cost less in money and attention.
