Risk and health
Risk levels, health, rarity and signals for each component.
Argunis/Intelligence/Cerebro
The failure, before it happens.
Cerebro learns each component’s normal behaviour, detects what is odd and estimates the risk of failure in the next hour, 6 and 24 hours. It warns you in time to prevent it.
Cerebro · 148 components
High risk3
Medium risk9
Precision87%
Risk in 6 hours
Forecasts with their range
On each component’s page you see its 7-day forecast against what actually happened, the variables that weigh most and the estimated time it will cross the threshold.
- Risk at 1, 6 and 24 hours
- 7-day forecast against reality
- Variables that weigh most
- Servers, web, cloud and APM services
Disk usage · db-salesFull on Thursday 03:10
Scenarios that repeat
Cerebro stores the state before every outage. When that scenario shows up again, it warns you with the cause from last time, so you act before it repeats.
- State before every outage
- Warning when the scenario repeats
- The cause from last time
- Priority in the SIEM inbox
Precursor scenario · api-payments89% match
- 1Connection pool at 95%
- 2Wait queues
- 3Latency ×3
- 4Outage (Sep 12)
Last time the cause was a deployment with a connection leak. Cerebro warns at step 2.
It learns from your answers
Every “may fail” alert is answered with how it turned out. With that, each component type’s model is validated and retrained, and you decide how many alerts each one may raise per week.
- Answer to each predictive alert
- Models per component type
- Validation and retraining
- Weekly alert cap and email warning
Alert precision · 12 weeks
Everything Cerebro includes
Anomaly mode
What departs from normal behaviour, even if it crosses no threshold.
Summary
How many components it watches, how many are at risk, how precise its alerts are and how it learns.
Devices
Every component with its risk at 1, 6 and 24 hours, its trend and its signals.
Component page
7-day forecast against reality, top variables and alerts.
Predictive alerts
The “may fail” alerts, with your answer so the model learns.
Models
Status of each component type’s model, its validation and next training.
Settings
Alerts per week, the horizon shown first and email warning at high risk.
Precursor scenarios
The state before each outage and a warning when it repeats.
InfrastructureFrequently asked questions
Which components can it predict?
Servers, web services, cloud services and APM services. It is turned on from each one’s page.
Does it need a lot of history?
It learns from each component’s behaviour and improves over time. Your answer to each alert speeds up that learning.
Does Cerebro close or hide alerts?
No. It only predicts and prioritizes; it never closes or hides an alert or a detection.
See the failure before your customers do.
We show you Argunis on your own operations and answer your questions in one call.