CloudWatch

• CloudWatch Metrics – services send time-ordered data points to CloudWatch

• CloudWatch Alarms – monitor metrics and initiate actions

• CloudWatch Logs – centralized collection of system and application logs

• CloudWatch Events – stream of system events describing changes to AWS resources and can trigger actions

Architecture

Data

Alarms

Two types of alarms

• Metric alarm – performs one or more actions based on a single metric

• Composite alarm – uses a rule expression and takes into account multiple alarms

• Metric alarm states:

  • OK – Metric is within a threshold

  • ALARM – Metric is outside a threshold

  • INSUFFICIENT_DATA – not enough data

Data Architecture

Amazon CloudWatch Metrics

• Metrics are sent to CloudWatch for many AWS services

• EC2 metrics are sent every 5 minutes by default (free)

• Detailed EC2 monitoring sends every 1 minute (chargeable)

• Unified CloudWatch Agent sends system-level metrics for EC2 and on-premises servers

• System-level metrics include memory and disk usage

• Can publish custom metrics using CLI or API

• Custom metrics are one of the following resolutions:

  • Standard resolution – data having a one-minute granularity

  • High resolution – data at a granularity of one second

• AWS metrics are standard resolution by default