Managing a Kubernetes cluster with 10 nodes? Easy. But what about 2,000? Scaling efficiently while handling diverse workloads—GPUs, batch jobs, stateful apps—becomes a real challenge. Without a solid strategy, costs spiral, and scheduling turns into chaos.

In this talk, you'll learn practical strategies for managing Karpenter at scale: optimizing node types, handling drift, deploying safely, and monitoring effectively. Whether you're new to Karpenter or already pushing its limits, this session will give you the insights needed to run smarter, not harder.

Ido Ziv - DevOps Team Leader at Kaltura, living the conflict of Best Practices and Fast Delivery. Currently leading the Infrastructure and DevEx field. Ido has a lot of passion and experience in Platform engineering, Infrastructure as code, GitOps and agile methodologies. He believes that the secret ingredient for creating a robust scalable systems is good design practices, effective communication between teams, and lots of automations. On his days off he likes sailing, did we say Kubernetes already?

Adi Ziv - Senior DevOps Engineer at Kaltura, specializing in cloud-native architectures and Kubernetes scaling. With over seven years of experience, he has led initiatives in automation, cost optimization, and infrastructure resilience. Adi is passionate about building scalable, event-driven systems and has implemented Karpenter to enhance Kubernetes efficiency. He believes that intelligent automation and dynamic resource management are key to reducing operational overhead. When not optimizing cloud workloads, he enjoys mentoring engineers and navigating the open sea—because scaling clusters isn’t so different from sailing!

Would you like to increase the performance and efficiency of your Kubernetes cluster while reducing its costs? Join us to learn how Karpenter, an open-source Kubernetes autoscaler, can help you achieve these goals. In this meetup, we'll explore ways to optimize the efficiency of your workloads and better utilize your existing resources. Through real-world examples we'll show you how we successfully run our workloads on 100% spot instances, even during high-demand periods. We will share our experiences implementing Karpenter in production environments, discuss common challenges we faced, and present the solutions that helped us achieve substantial cost savings while maintaining reliable operations.

Rotem Maayan - Staff DevOps Engineer at Unity with over a decade of experience in senior DevOps and cloud infrastructure roles. I specialize in architecting and operating large-scale Kubernetes and cloud environments, especially on AWS and GCP. Beyond my technical skills, I practice yoga and embrace stepping out of my comfort zone to continuously evolve both personally and professionally.

In this session, I will discuss how we initially adopted Karpenter at Wix to streamline our node replacement process, eliminating the complexity of lifecycle hooks and reducing the need for manual intervention. Once we successfully integrated it for this purpose, we broadened our evaluation to assess its scalability, cost efficiency, and real-time provisioning capabilities.

Our POC included extensive stress testing to validate Karpenter’s performance under heavy load and ensure seamless integration with our Kubernetes environment. Along the way, we encountered challenges with node termination logic, fine-tuning scaling behavior, and monitoring integration. However, we recognize Karpenter’s potential and are now focused on optimizing our setup and exploring additional features to further enhance our infrastructure.

Evyatar Shtern is a DevOps Engineer with extensive experience in designing, automating, and managing infrastructure across multiple companies. I have a strong background in system administration, cloud technologies, and process optimization, with a focus on scalability and reliability.