About shape correlation
Shape correlation is an alert grouping mechanism that analyzes metric data to automatically identify and correlate time series that exhibit similar patterns or shapes. This correlation occurs across different resources, domains, and data sources without requiring manual configuration.
Shape correlation addresses the common operational challenge of identifying related issues across your infrastructure when multiple resources exhibit similar metric patterns. For example, when a host experiences unusual CPU wait-time spikes and I/O spikes, shape correlation can automatically identify which containers or services running on that host show similar metric patterns, helping you quickly pinpoint the root cause. It can handle millions of time series without performance degradation.
How shape correlation works
The shape correlation algorithm continuously analyzes metric data to detect time series that have similar shapes or patterns. When time series from different resources show similar behavior over time, they are automatically grouped together as related alerts. The shape correlation algorithm exhibits the following features:
- Operates across different resource types and data sources
- Requires no additional hardware or manual configuration
- Scales efficiently regardless of the number of time series being analyzed
- Works in conjunction with other correlation mechanisms (temporal, topological, and scope-based)
Benefits of shape correlation
Shape correlation provides several operational advantages:
- Efficiency gains:By grouping related alerts based on metric patterns, operations teams face fewer tickets and reduced alert noise. It also reduces duplication and improves cross-domain visibility.
- Improved context: Seeing correlated issues together based on their metric behavior helps teams diagnose problems faster and understand the relationships between different components.
- Impact awareness: Understanding that an alert is part of a broader pattern of similar metric behavior aids in prioritization and root cause analysis.
Examples of shape correlation
Consider two related metrics from a Java application:
- Garbage Collection Rate (GC_Rate)
- Garbage Collection Times Run (Times_Run)
When displayed on a normalized chart, these time series may appear nearly identical in shape, even though their actual values differ. Shape correlation automatically identifies and groups these related time series, indicating they are part of the same underlying issue.
Similarly, in the following example, multiple time series with similar shape are correlated. This ensures that you have one problem for one person, rather than multiple tickets for multiple people to look at.
