-
Open Infrastructure in Europe: What Comes Next? | Fireside Chat with Thierry Carrez
Thierry Carrez
-
Ankommen & WarmUpOkt.. 24
-
BegrüßungMatthias TinnemeierOkt.. 24
-
Keynote - Effective Control: Going Beyond Digital Sovereignty
Thierry CarrezOkt.. 24
-
How to Build a Central Storage Platform for Research and Administration Today?
Andreas NiederländerOkt.. 24
-
Predicting GitOps at 15,000+ Clusters: What Large-Scale Testing Taught Us
Artem LajkoOkt.. 24
-
Building a Scalable Alternative to Hyperscale Clouds
Christoph StreitOkt.. 24
-
Benchmark vs Test – OpenStack Rally in Practice
Michel RaabeOkt.. 24
-
Automatic Eviction and Migration of VMs in OpenStack Clusters Running on k8s
Stefan HoffmannOkt.. 24
-
OpenStack backup aligned: multitenancy, data and metadata
Artur WojcikOkt.. 24
Tobias Weiss is a PhD in Business Informatics (Wirtschaftsinformatik) with a passion for Cloud, AI, and DevOps. He combines academic foundations with hands-on implementation: from infrastructure automation and cloud architecture to agentic AI workloads.
In his daily work he builds and operates sovereign cloud infrastructures — including an SCS cluster (3-node OpenStack + OVN + Ceph) on reused hardware, where he diagnosed and pragmatically worked around an OVN 25.03.1 gateway redirect bug. His motto: "Not macht erfinderisch" — necessity breeds invention; constraints are opportunities for clever solutions.
As an innovator (from AI to XR) and entrepreneur, he connects cloud infrastructure with the next generation of workloads such as openDesk Edu — from research (tobias-weiss.org) and graph-based AI systems (graphwiz.ai) to practical AI literacy training (ki-kompetenz-training.org).
What if your cloud's constraints became its superpower?
In this practical case study, we run openEduSuite — a sovereign, AI-powered educational platform — on a 3-node SCS-verified bare-metal K3s cluster built from legacy Huawei servers, the kind of hardware most teams would recycle.
We walk the full stack: Huawei hardware repurposing, a lean bare-metal K3s platform, Ceph storage — and the agentic layer on top: a predictive agent doing Kalman + Markov risk scoring for proactive cluster maintenance, alerting before outages instead of after.
Along the way we share the real fights: an earlier OpenStack/OVN phase collapsed on a 25.03.1 gateway redirect bug that broke floating IPs. We ditched the VMs, moved the workloads to SCS-verified bare metal (6/6 KaaS standards PASS, 444/444 CNCF conformance), and made quantization plus deterministic scheduling compensate for the missing GPU.
The theme: "Not macht erfinderisch" — necessity breeds invention. Constrained hardware and upstream bugs forced pragmatic engineering — the result is a reproducible, documented setup.
As a workshop element, you install opencode on your own laptop, connect it to the free Big Pickle model, and run your first prompts against real cluster documentation — sovereign AI hands-on, no API key, no cost.
Slides, runbooks and compliance artifacts:
https://github.com/tobias-weiss-ai-xr/sovereign-cloud-days-2026
