Amazon CloudWatch Metrics and Dashboards

Amazon CloudWatch is the central monitoring and observability service in AWS. It collects metrics, logs, and events from virtually every AWS resource and allows you to visualise them in dashboards, set alarms, and take automated actions. In this lesson we focus on the two foundational building blocks: metrics and dashboards.

---

What Is a CloudWatch Metric?

A metric is a time-ordered set of data points that represents a variable you want to monitor — for example, the CPU utilisation of an EC2 instance or the number of messages in an SQS queue.

Every metric is defined by three pieces of information:

Component	Description	Example
Namespace	A container that groups related metrics	AWS/EC2, AWS/RDS, Custom/MyApp
Metric name	The measurement being tracked	CPUUtilization, RequestCount
Dimensions	Key-value pairs that identify the specific resource	InstanceId=i-0abcdef1234567890

Standard vs Custom Metrics

Standard metrics are published automatically by AWS services at no extra cost. For example, every EC2 instance emits CPUUtilization, NetworkIn, NetworkOut, DiskReadOps, and more — all at five-minute granularity by default.

Custom metrics are metrics you publish yourself using the CloudWatch API, CLI, or an SDK. Common use cases include:

• Application-level latency
• Business KPIs such as orders per minute
• Memory utilisation on EC2 (not included in standard metrics)

Publishing a custom metric is straightforward with the AWS CLI:

aws cloudwatch put-metric-data \
  --namespace "Custom/MyApp" \
  --metric-name "OrdersProcessed" \
  --value 42 \
  --unit Count

Metric Resolution

Resolution	Data Point Interval	Cost
Standard	5 minutes	Free for AWS-published metrics
High resolution	1 second	Additional charge

High-resolution metrics are useful for latency-sensitive workloads where five-minute averages hide important spikes. You enable high resolution by setting the StorageResolution parameter to 1 when publishing custom metrics.

---

Metric Statistics and Periods

When you query a metric, CloudWatch aggregates the raw data points over a period using a statistic:

Statistic	What It Returns
Average	Mean of data points in the period
Sum	Total of all data points
Minimum	Lowest value in the period
Maximum	Highest value in the period
SampleCount	Number of data points
pN (percentile)	The value below which N % of data points fall, e.g. p99

Choosing the right statistic matters. For latency metrics, Average can be misleading because a few very slow requests get hidden by thousands of fast ones. Use p99 or p95 to understand the experience of your slowest users.

---

Metric Retention

CloudWatch retains metric data according to this schedule:

Data Point Interval	Retained For
< 60 seconds (high-resolution)	3 hours
60 seconds	15 days
5 minutes	63 days
1 hour	455 days (15 months)

Older data is automatically rolled up into coarser granularity. If you need to keep high-resolution data longer, export it to S3 or a time-series database.

---