Quick question for the cost nerds: which of the big three launched per-second VM billing first, and what year? I remember trimming idle minutes off EC2 bills around late 2017, and it changed how I right-size autoscaling to squeeze out waste. Anyone have the exact rollout date and whether Azure or GCP beat them to it?
AWS went first: announced per‑second billing Sep 18, 2017 (effective Oct 2) for Linux EC2 and EBS; GCP followed on Sep 26, 2017 with per‑second after a one‑minute minimum, and Azure was still per‑minute then. From experience, when I switched I cut ASG cooldowns to about 60s so short spikes wouldn’t pay extra beyond the one‑minute floor. Source: New – Per-Second Billing for EC2 Instances and EBS Volumes | AWS News Blog.
One small gotcha from Oct ’17: AWS’s change only applied to Linux EC2 (plus EBS/EMR), not Windows/RHEL/SUSE, and some services (e.g., ELB/ALB) still had hourly components. My tip: split ASGs by OS and bias scale‑in toward Linux; after I did that, the “trim idle minutes” effect you mention got noticeably bigger. Did you have mixed fleets then, or all Linux?