News
Amazon CloudWatch Adds Instance-Level EBS I/O Limit Checks for Faster EC2 Troubleshooting
Amazon Web Services (AWS) announced two new Amazon CloudWatch metrics that show when an application is exceeding the EBS I/O performance limits of its EC2 instance. The metrics, named Instance EBS IOPS Exceeded Check and Instance EBS Throughput Exceeded Check, report a per-minute status so operators can see if driven IOPS or throughput are above the maximum that the attached instance can sustain. Each metric returns 0 when performance has not been exceeded and 1 when it has, providing a simple signal for dashboards, alarms, and automated remediation.
These metrics are enabled by default at a 1-minute frequency and are available at no additional charge, AWS explained. They apply to EBS volumes attached to Nitro-based EC2 instances across all Commercial AWS Regions, including AWS GovCloud (US) and China Regions. Teams can view the data in the EC2 console, query it via the CLI, or integrate it through the CloudWatch API. Because the output is normalized to a binary state, it is straightforward to place on CloudWatch dashboards and to trigger alarms or runbooks that scale up, shift instance families, or otherwise address storage throttling conditions.
The Instance EBS IOPS Exceeded Check indicates whether the observed read and write operations per second on a volume are surpassing the instance's EBS-Optimized limit. The Instance EBS Throughput Exceeded Check performs the same role for sustained bytes per second. Used together, they help distinguish whether an application slowdown is due to hitting instance-level EBS ceilings rather than volume configuration or application behavior. This distinction is useful during load tests and incident response, where minute-by-minute visibility can guide right-sizing decisions for instance type or size.
Operationally, customers can add these checks to existing monitoring alongside volume-level metrics. A common pattern is to create CloudWatch alarms that alert when either metric remains at 1 over a defined window, indicating persistent saturation. From there, responders can consult documented EBS-Optimized limits for the running instance type and decide whether to migrate to a larger size, change families, or adjust workload characteristics.
About the Author
David Ramel is an editor and writer at Converge 360.