Anomaly detection

AI automatically detects and prioritizes problems

What is anomaly detection?

Anomaly detection is a technique that uses AI to identify abnormal behavior as compared to an established pattern. Anything that deviates from an established baseline pattern is considered an anomaly. Dynatrace’s AI autogenerates baseline, detects anomalies, remediates root cause, and sends alerts.

A deterministic AI-powered approach to anomaly detection

Eliminate alert floods and alarm spam

Are your teams getting fatigued with alarms and alert storms? Dynatrace automatically connects related performance issues into more manageable alerts. Less noise, more problem-solving.
Focus on the problems that really matter

Not every threshold violation is a problem, and not all problems are serious. Davis AI accurately classifies whether an anomaly is serious and has an actual or potential impact on customers.
Accurate thresholds with dynamic baselining

Dynatrace identifies accurate thresholds for anomaly detection using sophisticated multidimensional baselining techniques. More accurate thresholds mean more accurate anomaly detection.

Anomaly detection built for dynamic environments

The traditional reactive approach of identifying problems by responding to alerts based on static thresholds doesn't work for today's dynamic, multicloud architectures with containers and microservices. What’s “normal” behavior is constantly redefined with containers spinning up and down and dependencies shifting between individual entities. These dynamic environments demand a new, proactive approach.

That's where AI comes in. With the deepest possible knowledge of your system's real-time topology, multidimensional baselining, and dynamic dependency detection, Dynatrace harnesses predictive analytics and continuous machine learning to automatically identify anomalies based on the metrics that really matter.

Detect unknown unknowns while avoiding alert storms and false positives triggered by static thresholds. End-to-end monitoring finds even the hardest-to-spot anomalies.

Dynatrace auto-detects all dependencies

Through smart auto-baselining and predictive analytics, Dynatrace automatically detects anomalies, and AI determines whether they impact customers

Dynatrace diagnoses problems that are often difficult to pinpoint, including database, memory, threading, and CPU issues

With Dynatrace, we have shortened the time to identify and solve performance problems by 60%, and have achieved 100% application performance visualization.

Yunpeng Qiao Senior Manager, Global Application Operation Lenovo

Prioritize problems automatically

Identify performance anomalies before they affect customers. Eliminate guesswork and stop spending time hunting down problems. Dynatrace automatically applies AI algorithms to determine whether a performance issue has an actual or potential impact on customers.

Because the anomaly detection engine understands the relationship between operational and business metrics, you get fewer meaningful notifications that tell you what and how customers' user experience is impacted.

Focus on fixing problems, not finding them.
Problem detection based on 100% of customer transactions—no averages or samples.

Smart baselining and prediction-based anomaly detection

Dynatrace uses different methodologies to determine when anomalous behavior warrants a problem notification. Automatic multidimensional baselining detects violations of individual reference values that change over time (response times and error rates of application or services). Predictive analytics detect abnormalities in application traffic and service load—as traffic and load depend on business-model seasonal patterns (e.g., workweek vs. weekends, day vs. night, Black Friday).

Dynatrace learns application traffic patterns and raises a problem notification when a statistically relevant deviation is detected—including a quantified customer impact and insight into the possible root cause.
Analytics can predict upcoming traffic levels and get smarter and smarter over time.
Automatic baselining can be fine-tuned for parameterized anomaly detection—lower thresholds for certain mission-critical services, or higher thresholds for apps and services still in development.