ML-Bound Anomaly Detection: Experiences Beyond Commodity Observability | Sumana Rao (EN)
Observability is an all-encompassing umbrella of a multitude of services and analysis aimed at making critical cloud-native applications maintainable by mitigating the risk of customer incidents. Key to this is to build a robust monitoring and alerting system such that it detects any anomalies in the system and notify the same to the concerned stakeholders at the earliest. But when should we trigger an alarm to the stakeholders? What constitutes a deviation or an anomaly? These are generally based on “accepted” levels, but this is a matter of debate.
We use Machine Learning (ML) to answer some of these questions – to identify patterns of “acceptable” behaviours, dynamically decide the threshold to flag off alarms and so on. Additionally, since the training data is different for different scenarios – for example, resource/memory consumption or network traffic in Kubernetes infrastructure, the model evolved takes care of arriving at tailor-made threshold levels. And the best part is, this matures over time. As practitioners, we use ELK (Elastic Stack with Kibana) as the primary tool for logging management and notification which offers anomaly detection exploiting its ML-enabled capability.
This session would provide pragmatic insights into ML-based Observability and Anomaly Detection capabilities using ELK that makes critical cloud-native applications maintainable by mitigating the risk of customer incidents.
➜ Abonniere JAVAPRO und verpasse keine News:
https://bit.ly/YouTube-JAVAPRO-abonni...
▬ Über uns: ▬
JAVAPRO ist das Magazin für professionelle Java-Entwicklung in der Praxis.
Über das Magazin hinaus ist JAVAPRO Veranstalter der jährlich stattfindenden Entwickler-Konferenz JCON. Auf der JCON stehen Core Java, Enterprise Java, Microservices, APIs und Frameworks im Fokus.
Auf diesem Kanal finden sich Interviews, Sessions, Tipps & Tricks und alles, was das (Java)-Entwickler-Herz begehrt.
➜ Jetzt kostenlos Magazin anfordern:
https://javapro.io
▬ Folge uns auf Social Media: ▬
Twitter: / javapromagazin
Facebook: / javapromag
➜ Danke fürs Abo & fürs Zusehen!