Despite widespread adoption of public cloud architectures in large enterprises, there are still use cases that are not suitable for a cloud-based solution for various technical or even legal reasons. These include, in particular, edge computing applications in which, due to high data volumes, transfer to the cloud is often not possible within the given time frame, and processing must therefore be carried out where the data occurs. At BMW this can result in several petabytes per use case per year. For such use cases, BMW has built a new edge computing platform that is designed to process data as close as possible to the source. Due to the use of this platform at a large number of different locations worldwide, the platform is especially designed for efficient operation and easy scalability. For example, central monitoring and provisioning is possible, i.e. new instances of the platform can be rolled out fully automatically using an Infrastructure-as-Code approach. In the presentation we will introduce the overall architecture of the solution and go into more detail on various topics. The listener learns how BMW has solved the following aspects:

• Fully automated setup of clusters on on-premise hardware
• Operating concept for a large number of clusters
• Self-service for platform services
• Multi-Cluster-Management
• Centralized logging and monitoring