CloudWatch is the monitoring service baked into AWS, and understanding what it does and where it stops explains most AWS observability decisions.
Amazon CloudWatch is AWS's native monitoring and observability service. It collects metrics, logs and events from AWS services and your own applications, lets you visualise them in dashboards, and triggers alarms and automated actions when thresholds are crossed.
CloudWatch has several parts. It collects metrics, every AWS service publishes them into a namespace, and you can add custom metrics, and stores logs through CloudWatch Logs, queryable with Logs Insights. Alarms watch metrics and fire actions (an SNS notification, an Auto Scaling adjustment) when a threshold is breached, and EventBridge routes events to targets. On top sit dashboards and higher-level features like Container Insights for ECS and EKS, Lambda Insights, Synthetics for scripted checks, and CloudWatch RUM for the browser. Because it is native, an AWS service starts publishing to CloudWatch with no setup, which is a large part of its appeal.
CloudWatch's strength is that it is native and zero-setup: every AWS service reports to it, it integrates with IAM and the rest of AWS, and you pay only for what you use. Its limits are the flip side of that. It is AWS-centric, so a hybrid or multi-cloud or on-premises estate falls outside it. Its default metric granularity is coarser and higher resolution costs more. And its pricing, per metric, per log gigabyte, per dashboard, per API request, adds up in ways teams often discover on the bill. Most importantly, like any monitoring service, it alerts a human; it does not resolve the incident.
CloudWatch and OpenTelemetry interoperate. The AWS Distro for OpenTelemetry (ADOT) can send telemetry into CloudWatch and X-Ray, and the OpenTelemetry Collector's awscloudwatch receiver can pull CloudWatch metrics into a portable, vendor-neutral pipeline. A common pattern is to keep CloudWatch for AWS-native metrics and use OpenTelemetry so that data can be correlated with the rest of your estate and sent wherever you want, rather than being locked to one cloud's console.
Ops Singularity ingests CloudWatch metrics over OpenTelemetry (via the awscloudwatch receiver and ADOT) alongside the rest of your estate, and Sentinel AI resolves the incidents they reveal, a throttled database, an unhealthy target, through governed Action Tickets, so AWS-native signals lead to correlated resolution rather than an isolated console.
CloudWatch has a free tier and is otherwise pay-as-you-go, priced per metric, log gigabyte, dashboard and API request. At scale, those charges add up, which is one reason teams evaluate alternatives.
CloudWatch is AWS-focused. For hybrid, multi-cloud or on-premises estates you need a tool that spans all of them, which is where OpenTelemetry and a cross-environment platform come in.
CloudWatch is AWS's monitoring service and backend; OpenTelemetry is a vendor-neutral standard for producing and moving telemetry. They interoperate through ADOT and the Collector's CloudWatch receiver.
Ops Singularity turns open telemetry into autonomous, governed resolution. See it on your own stack.