Server monitoring company Datadog today added an anomaly detection feature to its software-as-a-service (SaaS) platform to identify servers behaving abnormally compared to their peers. In addition, the company has upgraded its diagnostics for Amazon Web Services’ cloud container service.

Datadog built its new anomaly detection features directly from its acquisition of Mortar Data in February. The features allow developers to customize input parameters and do analysis across all hosts without having to set a fixed threshold for what is considered normal versus anomalous. When a host appears to deviate from the rest, it is automatically flagged as a rogue server.

The anomaly detection is powered by two algorithms that analyze scaling data streams with a single query: DBSCAN and median absolute deviation (MAD). DBSCAN is a clustering algorithm that aggregates host points close to one another. Clusters with few points above or below a threshold are considered outliers. The MAD algorithm calculates variability across data.

Datadog says its new monitoring capabilities for Amazon ECS (Amazon’s container management service) were driven by rapid customer adoption of both AWS and Docker.

Enhancements to Datadog’s platform for Amazon ECS include:

• Visibility into the health of a Docker-powered service
• Infrastructure slicing for visibility into resource usage, response times, and errors
• Automatic tracking of containers as they come online or terminate
• Alerts about any Docker-powered service under-performance.

Datadog, founded in 2010, raised a $31 million Series C in January, bringing its total funding to $52.2 million. A spokesperson for the company says Datadog displaces enterprise monitoring platforms like IBM Tivoli and BMC Patrol.