Amazon CloudWatch provides a general purpose timeseries database used to monitor resources and applications in real-time, collecting and tracking metrics and log messages for use in dashboards or alarms and delivering notifications when alarms are raised.
- Metrics are data points we're monitoring. They can be monitored at various frequencies, which can be adjusted as required (at cost).
- Alarms can trigger behaviour to remediate problems or alert operators based on metrics falling out of defined thresholds.
- Events deliver real-time streams of events taking place in a service or on a regular basis, allowing us to respond.
- Dashboards combine metric visualisations with explanatory text and images.
Namespaces are containers for metrics, allowing isolation between applications.
|<60 seconds||3 hours|
|60 seconds (1 minute)||15 days|
|300 seconds (5 minutes)||63 days|
|3600 seconds (1 hour)||455 days (15 months)|
Higher resolution data points will be aggregated together for long-term storage, degrading over time per the table above.
Data points older than 15 months are removed on a rolling basis. They can't be deleted, but expire after 15 months if no additional data is pushed to them.
A dimension is a name-value pair that is part of a metric's identity and can be used to query it. Each metric can be assigned up to 10 dimensions.