Live Migration of PyTorch GPU Nodes From Azure To European Clouds

on YouTube ↗07m

" summary

Live migration of a full GPU PyTorch VM including CUDA drivers, kernels and Jupyter notebooks runs by exporting disks, transforming the OS image and importing into the target cloud, then booting a parallel VM so rollback is immediate. The demo uses DigitalNomadSky to move from Azure to Leaf Cloud and similar European providers, reports typical times of 5–30 minutes for real projects, and shows one government team halving costs, gaining more GPUs and resolving EU sovereignty/GDPR concerns.

AI-written summary · about

" related talks

• 32mGPU Machine Learning From Laptop to CloudMark Puddick · KubeCon + CloudNativeCon Europe 2019
• 59mBuild and deploy production ready PyTorch modelsHenk Boelman · NDC Porto 2022
• 41mA Future Journey: How to Migrate 100 Clusters Between Clouds Without Downtime?Tobias Schneck · KubeCon + CloudNativeCon North America 2020 Virtual
• 22mInstant provisioning and hot migrationGilbert Cabillic · Conf42 Cloud Native 2022
• 22mEffective migration from Newton to Queens with minimal downtimeAttila Szlovencsák, Viktor Schlaffer · OpenInfra Summit 2020
• 44mGetting started with Azure Machine Learning servicesHenk Boelman · Future Tech 2019
• 48mStreamlining your Machine Learning projects with Azure Machine LearningSam Vanhoutte · Build Stuff 2021
• 09mDemocratizing Access to GPUs for Faster MLAI Innovation with Generative AIDuncan Ng · Great International Developer Summit 2023