CloudWatch Metrics, Namespaces, and Dimensions

Amazon CloudWatch is AWS's unified observability service that collects metrics, logs, and traces from your AWS resources and applications. It gives you a complete picture of your infrastructure health, enabling you to detect anomalies, set alarms, and automatically react to operational changes. CloudWatch is deeply integrated with almost every AWS service and is the primary monitoring tool tested in the SAA-C03 exam.

A CloudWatch metric is a time-ordered set of data points representing the value of a measurable aspect of a resource over time — for example, CPUUtilization for an EC2 instance measured every minute. Metrics are identified by their namespace, name, and dimensions. AWS publishes hundreds of built-in metrics automatically, and you can publish your own custom metrics from your application code or scripts.

A namespace is a container for CloudWatch metrics that prevents name collisions between metrics from different sources. AWS services use namespaces like AWS/EC2, AWS/RDS, AWS/Lambda, and AWS/S3. When you publish custom metrics, you choose your own namespace — for example MyApp/OrderService. Metrics in different namespaces are completely isolated from each other even if they share the same metric name.

Dimensions are key-value pairs that uniquely identify a metric within a namespace. For example, in the AWS/EC2 namespace, CPUUtilization can be filtered by the dimension InstanceId=i-0123456789abcdef0 to see one specific instance, or by AutoScalingGroupName=my-asg to see the aggregated metric across a fleet. A single metric can have up to 30 dimensions and AWS pre-defines which dimensions are available for each service metric.

By default, AWS services publish metrics at 5-minute intervals (standard monitoring) to CloudWatch at no extra charge. For time-sensitive workloads you can enable detailed monitoring to get 1-minute granularity. Detailed monitoring is available for EC2, Auto Scaling, ELB, RDS, and other services — it incurs additional CloudWatch charges. For SAA-C03, remember that Auto Scaling policies using step scaling rely on CloudWatch alarms that work better with 1-minute metrics.

You can push custom metrics to CloudWatch from any application using the put-metric-data API or the CloudWatch Agent. Common custom metrics include business KPIs (orders per minute), application-level latency, queue depth, and error rates. Custom metrics are retained for 15 months with variable resolution, costing per metric per month. High-resolution custom metrics (1-second granularity) are also supported at higher cost.

The CloudWatch Agent is a software agent you install on EC2 instances (and on-premises servers) to collect metrics not available via the built-in EC2 hypervisor, such as memory utilisation, disk space, and network connections. The agent also collects logs and sends them to CloudWatch Logs. Memory utilisation is a common exam scenario — it is NOT a default EC2 metric and MUST be collected via the CloudWatch Agent.

Metric math lets you create new time series by performing arithmetic on existing metrics without storing the result as a separate metric. For example, you can compute error rate as Errors / Requests * 100 directly in a CloudWatch dashboard expression. Anomaly detection applies machine learning to historical metric data to create expected value bands, and you can create alarms that fire when a metric deviates outside those bands — even if there is no fixed threshold.

CloudWatch stores metrics with varying retention based on their resolution: data points at 1-second resolution are retained for 3 hours, 1-minute data for 15 days, 5-minute data for 63 days, and 1-hour data for 15 months. CloudWatch automatically aggregates high-resolution data into lower-resolution data over time. This means if you need to analyse long-term trends you will see 1-hour averages — not the original 1-minute readings from months ago.

The CloudWatch console provides a Metrics Explorer and Metrics browser where you can select any namespace, metric, and dimension combination and visualise data over time. You can overlay multiple metrics on a single graph, change aggregation periods, and add them to dashboards. For programmatic access, the GetMetricStatistics and GetMetricData APIs return raw data points or aggregate statistics (Average, Sum, Minimum, Maximum, SampleCount).

For the SAA-C03 exam, know these critical metrics: EC2 — CPUUtilization, NetworkIn/Out, DiskReadOps/WriteOps (instance store only); missing from EC2 by default: memory, disk space (use CloudWatch Agent). RDS — DatabaseConnections, FreeStorageSpace, ReadLatency, WriteLatency, ReplicaLag. ELB — RequestCount, TargetResponseTime, UnHealthyHostCount. Lambda — Invocations, Errors, Duration, Throttles, ConcurrentExecutions.

Test your understanding of AWS Solutions Architect (SAA-C03) concepts from this lesson.

In this lesson you learned: namespaces organise metrics by service and prevent name collisions, dimensions filter metrics to specific resources like a single EC2 instance or ASG, and memory utilisation is not a default EC2 metric — you need the CloudWatch Agent for that. Next up we explore CloudWatch Alarms and Composite Alarms for automated alerting.