Kedify

🚀 Putting Knowledge into Action: Our Team at #KubeConEU 2026! This week, Amsterdam became the global hub for cloud innovation at KubeCon + CloudNativeCon Europe. With over 15K attendees from across the globe, it is the ultimate stage for the open-source community, and our team was right at the center of it. We are incredibly proud to have had 3 expert sessions led by our own colleagues, sharing real-world insights on the technologies shaping the future: 🔹"From FTP To Argo CD & Argo Rollouts: An Adoption History Inside a Corp."  a masterclass on modernization by Alfonso Ming Cazorla & Meritxell Rodríguez Torrejón shared at ArgoCon Stage. 🔹"The Good, The Ugly, and The Bad: Leaving Sidecars Behind with Istio Ambient Mesh."  tackled the evolution of service mesh by Alfonso Ming & Jorge Turrado Ferrero at Istio Day 🔹"Unleashing Event Driven Capabilities With KEDA." by Jorge Turrado Ferrero that teamed up with Zbyněk Roubalík (Kedify) to dive deep into scaling at the main event. Participating in KubeCon isn't just about attending; it’s about contributing to the global conversation. By sharing our journey and collaborating with the best in the industry, we ensure we stay at the cutting edge of Cloud Native excellence. A massive shout-out to our team for showcasing our technical excellence on the global stage! 🌍🚀👏 #KubeCon #CloudNative #OpenSource #Kubernetes #ArgoCD #Istio #KEDA #TechLeadership

Zbyněk Roubalík, Founder & CTO @ Kedify, discusses the fundamental limitations of traditional Horizontal Pod Autoscaler (HPA) metrics and advocates for a shift toward predictive autoscaling approaches. He explains that CPU and memory metrics work well for stable workloads but become inadequate when dealing with variable workloads, particularly for HTTP traffic and gRPC services that require real-time responsiveness. Watch the full interview: https://ku.bz/vc-lBjCr0 This interview is a reaction to Brian's episode https://ku.bz/sFd8TL1cS

Zbyněk Roubalík, Founder & CTO @ Kedify, explains how Kedify's new features address gaps in the Kubernetes autoscaling landscape. He discusses three key innovations: predictive scaling (which he notes has no real competitive solutions beyond experimental Prometheus setups), LLM workload autoscaling using the vLLM open-source framework to properly scale AI inference workloads, and multi-cluster autoscaling that overcomes the single-cluster limitation of open-source KEDA. Watch the interview: https://ku.bz/qN7BLcYTK Read the announcement: https://ku.bz/0XVsNHSnK

Zbyněk Roubalík, CTO Founder @ Kedify, shares his perspective on effective auto-scaling strategies for Kubernetes. He breaks down auto-scaling into two essential levels: cluster-level scaling for nodes and application-level scaling. He stresses that successful auto-scaling requires choosing the right metrics beyond just CPU and memory to make scaling decisions. Watch the full interview: https://ku.bz/k0xRzNBVd This interview is a reaction to Thibault's episode https://ku.bz/rf1pbWXdN

Zbyněk Roubalík, Founder & CTO @ Kedify, explains the current limitations of Kubernetes cluster autoscaling and his vision for more intelligent infrastructure provisioning. He discusses how cluster autoscalers like Karpenter work well with pod-level pressure from HPA, scaling based on CPU and memory requests, but argues for a more proactive approach. Zbyněk highlights the technical challenge of connecting different types of metrics to enable proactive cluster scaling - a capability that's currently very difficult to achieve in Kubernetes. Watch the full interview: https://ku.bz/vc-lBjCr0 This interview is a reaction to Jorrick's episode https://ku.bz/clbDWqPYp

Zbyněk Roubalík, CTO Founder @ Kedify, discusses the critical considerations for implementing auto-scaling safely in production environments. He emphasizes that auto-scaling requires continuous monitoring and tuning, with constant attention to metrics and observability tools to track application performance. Watch the full interview: https://ku.bz/k0xRzNBVd This interview is a reaction to Thibault's episode https://ku.bz/rf1pbWXdN