Cloud Native Prague #30

Running Kubernetes on-premise comes with challenges that managed cloud solutions simply abstract away — and ingress traffic is one of them. In this talk, I want to walk through how we handle traffic ingestion at scale in our own datacenter at Livesport, using a two-tier load balancing setup that has proven itself under real production load. I'll cover the key design decisions behind our architecture, how we tie everything together to make scaling independent and predictable, and share the bumps we hit along the road.

Many teams are still overpaying for infrastructure due to overprovisioned resources, idle workloads, and reactive scaling strategies. Even with modern cloud tooling, it’s hard to continuously right-size systems as usage patterns shift and architectures become more distributed.

In this session, I’ll walk through a real-world use case of applying agentic AI to infrastructure optimization. Instead of relying on static rules or dashboards, we use autonomous agents to analyze telemetry, identify inefficiencies, and take corrective actions—such as rightsizing compute, tuning autoscaling policies, and eliminating waste across environments.

We’ll dig into how the system is built: how agents consume metrics and logs, how decisions are evaluated and executed safely, and how to integrate this approach with existing platforms (e.g., Kubernetes, cloud providers, observability stacks).

By the end, you’ll have a clear understanding of how to implement a similar approach in your own environment, what trade-offs to consider, and where agentic AI provides the most leverage compared to traditional optimization techniques.