AWS CloudWatch, Explained Honestly
CloudWatch is the monitoring AWS gives you with the account: metrics for every service, logs, dashboards, and alarms that can act on their own. Inside AWS it's unavoidable and mostly right. The judgment calls are about cost discipline — and about everything in your estate that doesn't have an ARN.
The Pieces of CloudWatch
Metrics
Native metrics from every AWS resource, plus custom metrics you publish.
Logs & Insights
Centralized logs with a query language for analysis.
Alarms & actions
Alarms that trigger SNS, auto-scaling, or Lambda automatically.
Agent for servers
OS metrics and logs from EC2 and on-prem machines.
Where CloudWatch Stops
CloudWatch's world is resources with AWS identities. The agent extends it to operating systems anywhere, but the physical layer below — server hardware health, BMC events, PDUs, thermals, rack capacity — is invisible to it by design. For hybrid estates the working pattern mirrors Azure's: CloudWatch owns the cloud half, an infrastructure platform like Sensaka owns the physical half down to component level, and events flow into one view so a database timeout in AWS and the failing SAN switch on-prem can finally meet in the same incident.
Common Questions
What is AWS CloudWatch?
CloudWatch is AWS's built-in monitoring service: metrics from every AWS resource, log collection (CloudWatch Logs), dashboards, and alarms that can trigger automated actions like auto-scaling or Lambda functions.
What does CloudWatch cost?
Pay-per-use across custom metrics, dashboards, alarms, log ingestion, and retention. Like most cloud-native monitoring, log volume is the line item that surprises — high-cardinality custom metrics come second.
Can CloudWatch monitor on-premises servers?
The CloudWatch agent can ship OS-level metrics and logs from on-prem machines into AWS. Below the OS — hardware health, BMC events, power, cooling, racks — CloudWatch has no visibility; that layer needs infrastructure-native tooling.
CloudWatch vs Azure Monitor — which is better?
Neither is portable: each is the right default inside its own cloud and mediocre outside it. Multi-cloud teams usually standardize dashboards elsewhere (Grafana, Datadog) and keep each native service for its own platform's depth.