Ref: https://learn.cantrill.io/courses/1820301/lectures/41301628

DEMO: https://learn.cantrill.io/courses/1820301/lectures/41301629

• Summary
• Contents:

Amazon CloudWatch - Components and Architecture

• 🔧 Collects & manages operational data → provides operational management & monitoring
  • Operation data includes service performance, metrics, logs…
  • Core support service used by almost all other AWS services
  • Public service

• Three main components:
  1. CW Metrics = "CW itself” (”Vanilla” CW)
     • e.g. CPU utilization of EC2 instance over time, disk usage on an on-premises server…
     • Can ingest metrics from AWS products, apps, or on-premises infrastructure
     • Some metrics are gathered natively by AWS products
     • Configuration of a CW Agent is needed for:
       • Collecting metrics inside AWS that are not gathered natively
         • e.g. internal processes in an EC2 instance
       • Collecting metrics outside of AWS
     • 💡 Lots of metrics, must be kept divided
  2. CW Logs
     • Can ingest logs from AWS products, apps, or on-premises infrastructure
     • Some logs generated natively by AWS products, other products need CW Agent
  3. CW Alarms
     • Trigger notifications (send to Amazon SNS) or events based on observed metrics
     • e.g. ring an alarm and send SMS when CPU usage of my EC2 instance surpasses 90%
     • 💡 The simple Billing Alarm we set before was set with CW → send a notification (email) when billing surpasses a budget
• ❗ NOTE: There used to be also CW Events, but its functionality is nowadays provided by Amazon EventBridge

Amazon CloudWatch - Key Concepts

• Datapoint = Timestamp + Value (Pair)
  • e.g. CPU usage is 6.7% at 10:32:67 on 2026-06-27
• Metric = Sequence of datapoints ordered in time = Time-ordered set of datapoints
  • e.g. CPU usage/utilization, NW I/O, Disk I/O… (in a given time frame)
  • ‼️ NOT necessarily related to a single server!
    • e.g. CPU usage will be displaying by default the usage across all EC2 instances! Must specify otherwise if you want to limit to one instance!
• Namespace = Name container for monitoring data, contains name-related metrics
  • 💡 Keeps things from getting messy
  • Can have any rule-complying name
  • e.g. AWS/ ← namespace where all AWS data goes to
    • e.g. AWS/EC2 ← namespace where all EC2 data goes to
• Dimensions = Criteria that separate datapoints of same metric into different perspectives
  • e.g. Inside AWS/EC2 namespace, separate into dimensions: Instance A, Instance B…
  • Powerful & Flexible
• Alarms take actions based on a specific metric reaching specified values/thresholds
  • e.g. Send a notification when budget has been exceeded (billing alarm)
  • States: INSUFFICIENT DATA (initial state), OK, or ALARM
    • alarm uses Amazon SNS (Simple Notification System) when the state is ALARM
  • Configurations to consider: period, evaluation period, datapoints to alarm