In a recent webinar, our Customer Success Manager, Graeme Smith, highlighted a powerful feature within FusionReactor Cloud: anomaly detection. This feature is designed to keep you informed about everything in your system, beyond traditional manual alerts.
Moving Beyond Manual Alerts
While manually configured alerts are useful, they can be time-consuming and require you to anticipate every potential issue. FusionReactor’s anomaly detection takes a proactive approach. It leverages machine learning to constantly monitor key performance indicators, which we call “RED metrics“:
- Rate of requests: How many requests are hitting your system?
- Rate of errors: How many errors are occurring?
- Duration of requests: How long are requests taking to process?
By analyzing these metrics, FusionReactor establishes a “normal baseline” and identifies deviations that could signal potential problems. This means you don’t have to guess what to monitor—the system intelligently watches for anything unusual.
How Anomaly Detection Works
Anomaly detection in FusionReactor Cloud uses machine learning to understand typical system behavior. It then alerts you when it detects significant deviations. You can even configure the sensitivity of these alerts, setting a threshold for the system’s safety before notifying you.
“I’ve said, look between now and a week ago. And if you’re 55% sure there will be an anomaly, please alert me.”
Alerting and Integration
FusionReactor Cloud provides flexible alerting options. You can receive email alerts or integrate with Slack for quicker notifications. “The email alerts will come to me at weekends, but during the week when I’m logged in, I want to know about them straight away. So it’ll tell me via Slack,” Graeme shared. While a direct Teams integration is in development, the Slack integration ensures you’re promptly notified of any anomalies.
Investigating Anomalies
When an anomaly is detected, FusionReactor Cloud provides tools for investigating it. You can view traces of all requests that occurred during the anomalous period, pinpointing the exact cause of the issue. This granular view allows you to identify errors and understand their context quickly.
For example, Graeme showed how the system identified a drop in request duration as a potential anomaly. While a lower duration might seem optimistic, it could indicate a problem, such as a site outage. The anomaly detection feature helps you see the bigger picture and investigate further.
Deeper Insights
FusionReactor Cloud doesn’t just tell you that there’s a problem, it helps you understand why. By examining error traces and request spans, you can gain deep insights into the root cause of performance issues. In one example, Graeme identified an error caused by a product that didn’t exist in the database table.
Beyond the Basics
The power of FusionReactor Cloud extends beyond basic monitoring. With integrations like OpenTelemetry (OTEL), you can bring telemetry data from other technologies, such as MySQL databases. This gives you a unified view of your entire technology stack within a single dashboard.
Key Takeaways on How to Detect Anomalies
- FusionReactor’s anomaly detection uses machine learning to identify unexpected performance issues.
- It monitors “red metrics”: rate of requests, rate of errors, and duration of requests.
- You can receive alerts via email and Slack, with Teams integration coming soon.
- The system provides tools to investigate anomalies and identify root causes.
- FusionReactor integrates with other technologies to give you a comprehensive view of your system.
By leveraging anomaly detection in FusionReactor Cloud, you can proactively address performance issues, reduce downtime, and ensure a smooth user experience.
