A VALUED CUSTOMER OF THIRDEYE DATA
Centerity: anomaly detection across more than ten thousand metrics
Centerity works with enterprise customers worldwide, monitoring and analysing the data their systems generate and detecting anomalies in it. Monitoring at enterprise scale produces a specific difficulty: the number of things being measured is large enough that a human cannot watch them, and rules written by hand cover only the failures somebody already thought of.
LOCATION
Boston, Massachusetts, United States
COMPANY SIZE
Mid-scale Enterprise
INDUSTRY
Information Technology › Explore solutions for this industry
AI Ops anomaly detection platform
ThirdEye Data built the anomaly detection platform that sits inside Centerity’s offering. It detects and predicts anomalies across server-side measures such as processor load, memory consumption and traffic in and out, and it does so across more than ten thousand metrics in a given dataset rather than across a handful chosen in advance.
Scale is what makes the approach different from rule-based monitoring. A threshold alert tells you when a number crosses a line somebody drew. It cannot tell you that a combination of measures, each individually normal, together looks like the shape that preceded the last outage. Detecting that across ten thousand metrics is only feasible if the system learns what normal looks like for each of them rather than being told.
The platform includes automated model selection, so it adapts to the characteristics of the data it is given rather than requiring a specialist to tune it per deployment. When it detects something it raises it in the dashboards and sends notifications, and it supports drilling down from the alert to the underlying cause, because an alert that says something is wrong without indicating where is a interruption rather than information.
It was built entirely on open source foundations, which matters for a product deployed into customer environments with their own licensing and security constraints.
Anomaly detection across more than 10,000 metrics in a single dataset, with automated model selection and root cause drill-down.
Related solution: anomaly detection and AI operations
Anomaly detection sounds like one problem and is really several. Detecting a spike is easy. Detecting a slow drift is harder. Detecting a combination of individually unremarkable measures that together signal trouble is harder still, and doing all three across thousands of metrics without producing so many alerts that people stop reading them is the actual engineering problem.
ThirdEye Data builds these systems for companies whose product is the detection, as at Centerity, and for companies detecting anomalies in their own operations. The design questions are the same either way: what counts as normal when normal changes over time, how confident the system needs to be before it interrupts someone, and how a person gets from the alert to the cause. We build the drill-down path deliberately, because the alert is not the deliverable. The explanation is.
Explore our anomaly detection and AI operations solutions ›
DO YOU HAVE A SIMILAR USE CASE?
Talk to our team to learn more about what we did for Centerity, and how that experience applies to your own environment.
Explore What We Did for Other Customers
