Amazon EC2 Instance
Does ZopNight manage Amazon EC2 Instance?
An EC2 instance bills per second (60-second minimum) at an hourly rate set by its instance type, whether or not anything uses it. ZopNight discovers instances every 6 hours, tracks 90 days of CloudWatch metrics, and stops non-production instances on schedule via ec2:StopInstances.
Rules that fire on Amazon EC2 Instance
At a glance
| Field | Value |
|---|---|
| Stop verb | ec2:StopInstances |
| Scheduling notes | instance stopped via ec2:StopInstances and restarted via ec2:StartInstances on schedule. |
Amazon Elastic Compute Cloud (EC2) provides resizable virtual servers in the AWS cloud. Instances are billed for every hour they run, whether or not anyone is using them, which makes idle development, test, and staging machines one of the largest sources of avoidable cloud spend.
The compute meter runs per second
An On-Demand instance accrues charges per second of running time, with a 60-second minimum, at a rate fixed by the instance type and region. Utilization never enters the calculation: an m5.2xlarge at 3 percent CPU costs exactly what it costs at 90 percent. Reserved Instances and Savings Plans lower the rate but never pause the meter. Commitment discounts only pay off on hours the instance was going to run anyway.
Stopping is cheap, not free
A stopped instance stops the compute meter, but its EBS volumes keep billing per provisioned GB-month and any public IPv4 address it holds keeps its own hourly charge. Termination is the only state that ends every meter. For non-production machines that must come back tomorrow, stopped is the right state: the residual storage cost is a small fraction of the compute rate.
Where EC2 spend goes wrong
Three patterns account for most of it. Development and staging instances run 168 hours a week to serve roughly 45 of them; nights and weekends are pure waste. Older instance families (m4, t2, c4) deliver less per dollar than current generations doing the same work. And oversized instances sit at single-digit CPU for months because nobody owns the resize decision.
How ZopNight works with EC2
Instances are discovered every 6 hours via AWS Resource Explorer 2 and the EC2 API. Hourly CloudWatch CPU, network, and disk metrics are kept with a 90-day lookback, joined by CWAgent memory and GPU metrics when the agent is installed. Per-instance cost comes from Cost Explorer or CUR 2.0. Schedules stop instances with ec2:StopInstances at the window close and start them with ec2:StartInstances at open; recommendations cover idle instances, rightsizing, Spot, and Graviton migration.
Finding your instances
EC2 console, then Instances in the left navigation. The view is region-scoped, so instances in other regions stay invisible until you switch the region selector. That is why console spot-checks miss so much of a fleet.