40% of a 350K Monthly Bill Was Idle: 6 Production Decisions from Tag Attribution to Cost-as-Code in FinOps
Overview At the end of the month, a CTO stared at a 350K RMB cloud bill and asked: “Where did all this money go?” Nobody could answer. 120 ECS instances, 3 K8s clusters, 50+ SLBs, scattered across two cloud accounts (Alibaba Cloud and Tencent Cloud). Finance only saw the aggregate bill. The engineering team only knew the instances they had created themselves. There was no unified view. Worse, after a full inventory, more than 40% of resources were running at under 30% utilization — idle ECS instances humming along, over-provisioned CPUs wasting cycles, Pods spinning empty during off-peak hours, and SLBs with no backend still billing by the hour....