Cloud Native Community Ruhr November Meetup

Welcome to the Cloud Native Ruhr Community Meetup! 👋👋

We are delighted to invite you to a new meetup (long time no see!), on November 5th, 2024. Join us to listen to great speakers and network with fellow Kubernetes enthusiasts over food and drinks! 🌭🍹🥨

Location: Edgeless Systems' HQ, Stadionring 1, 44791, Bochum.

Agenda:
6:00-6:20 Drinks and welcome
6:20-6:40 "A Deep Dive into the Kubernetes Kubelet" by Peter Rossbach (Founder, bee42)
6:40-6:50 Q&A
6:50-7:20 "Build & Run GPU workloads on a Cloud Native Environment" by Niclas Mietz (Platform Engineer, DeepHealth)
7:20-7:30 Q&A
7:30-8:30 Networking, Würstchen & Brotchen (Also vegan options!), refreshments (Beers, soft drinks)

Talks abstracts:
A Deep Dive into the Kubernetes Kubelet
Kubernetes thrives on the flexibility and modularity of its API implementations. Join the talk and we delve under the hood of a Kubelet to explore the various implementation options for CRI, CNI, CSI, CDI, and NRI. Through practical examples, we'll showcase use cases for diverse container runtimes, from process virtualization and virtual machines to WebAssembly runtimes and hybrids. Discover which network and storage implementations yield the best results in real-world scenarios, and understand their impact on security and performance.

Build & Run GPU workloads on a Cloud Native Environment
As GPU-based workloads become indispensable for modern AI/ML applications, managing them efficiently in a Kubernetes environment is key to unlocking their full potential at scale while minimizing resource waste. This session dives into the complete process of preparing Kubernetes nodes for GPU workloads, starting with the comparison between NVIDIA GPU Operator and GKE’s native GPU driver. We'll then explore how Node Feature Discovery enhances hardware awareness, allowing for optimized resource management. From there, we’ll cover the development of a robust CI/CD pipeline using GitLab, designed to handle GPU jobs seamlessly and at scale. Attendees will also gain insight into advanced GPU concurrency mode and the critical aspects of running GPU workloads in clusters. Whether you're deploying machine learning models or scaling GPU workloads across clusters, this talk will provide you with actionable tools and best practices to optimize GPU utilization and streamline operations in your Kubernetes environment.

Are you interested in speaking or contributing at the next Meetup? Send us your submission here.

This group is managed and sponsored by Edgeless Systems, a cybersecurity start-up based in Bochum, Germany. Edgeless Systems develops leading open-source software for confidential computing. Find more info here.