AWS Architecture Blog

Developing a Platform for Software-defined Vehicles with Continental Automotive Edge (CAEdge)

This post was co-written by Martin Stamm, Principal Expert SW Architecture at Continental Automotive, Andreas Falkenberg, Senior Consultant at AWS Professional Services, Daniel Krumpholz, Engagement Manager at AWS Professional Services, David Crescence, Sr. Engagement Manager at AWS, and Junjie Tang, Principal Consultant at AWS Professional Services.

Automakers are embarking on a digital transformation journey to become more agile, efficient, and innovative. As part of this transformation, Continental created Continental Automotive Edge (CAEdge) – a modular multi-tenant hardware and software framework that connects the vehicle to the cloud. Continental collaborated with Amazon Web Services (AWS) to develop and scale this framework.

At this AWS re:Invent session, Continental and AWS demonstrated the new and transformative vehicle architectures and software built with CAEdge. These will provide future vehicle manufacturers, Original equipment manufacturers (OEMs) and partners with a multi-tenant development environment for software-intensive vehicle architectures. These can be used to implement software, sensor and big data solutions in a fraction of the development time needed before. As a result, vehicle software can be developed and tested more efficiently, then securely and rolled out directly to vehicles. The framework is already being tested in an automotive manufacturer’s series development.

Addressing core automotive industry pain points

Continental, OEMs and other major Tier 1 companies are required to quickly adapt to new technology requirements without knowing capacity or scaling needs, while at the same time staying ahead of the market. Developers are facing several challenges, in particular the processing of huge amounts of data. For example, a single test vehicle for AV/ADAS generates 20 – 100 TB of data per day. The handling of these data sets and the time to availability in distributed sites can cause major delays in development cycles. Delays are also experienced by developers due to the high numbers of test cases in simulation and validation. In an on-premises environment, this poses significant costs and scaling challenges to provide the required capacity.

The pace of the required transformation to becoming a software-centric organization is creating new challenges and opportunities like:

  • Current electronic architectures are decentralized, expensive, and complex to develop therefore difficult to maintain and extend.
  • Vehicle and cloud converge require new software (SW)-defined architectures, integration and operations competencies.
  • Digital Lifecycle Management enables new business models, go- to-market strategies and partnerships.

In addition to the distribution of huge datasets and distributed work setups is a need for cutting edge security technologies. Encryption at transfer/rest, data residency laws, and secure developer access are common security challenges and are addressed using CAEdge technology.

In this blog post, we describe how to build a secure multi-tenant AWS environment that forms the foundation for CAEdge. We discuss how AWS is helping Continental build the base infrastructure that allows for fast onboarding of OEMs, partners and suppliers through a highly automated process. Development activities can start within hours, instead of days or weeks; with a bootstrapped development environment for software-intensive vehicle architectures. This is all while meeting the strictest security and compliance requirements.

Overview of the CAEdge Framework

The following diagram gives an overview of the CAEdge Framework:

Architecture Diagram showing the CAEdge Platform

Figure 1 – Architecture Diagram showing the CAEdge Framework

The framework is based on the following modular building blocks:

  • Scalable Compute Platform – High Performance, embedded computer with automotive software stack and connection to the AWS cloud.
  • Cloud – Cloud services for developers and end-users.
  • DevOps Workbench – Toolchain for software development and maintenance covering the entire software lifecycle.

The building blocks of the framework are defined by clear API operations and can be integrated easily for various use cases, such as different middleware or CI / CD pipelines.

Overview of the CAEdge Multi-Tenant Framework

Continentals’ core architecture and terminology for a vehicle software development framework include:

  • CAEdge Framework as an Isolated AWS Organization – Continental’s CAEdge framework runs in a dedicated AWS Organization. Only CAEdge-related workloads are hosted in this AWS Organization. This ensures separation from any other workloads outside of the CAEdge context. The CAEdge framework provides multiple central security, access management, and orchestration services to its users.
  • Isolated Tenants – The framework is fully tenant-aware. A tenant is an isolated entity that represents an OEM, OEM sub-division, partner, or supplier. A key feature of this system is to ensure complete isolation from one tenant to another. We use a defense-in-depth security approach to ensure tenant separation.
  • Tenant-Owned Resources and Services – Each tenant has a dedicated set of resources and services that can be consumed and used by all tenant users and services. Tenant-owned resources and services include, but are not limited to:
    • Dedicated, tenant-specific data lake,
    • Tenant specific logging, monitoring, and operations,
    • Tenant-specific UI.
  • Projects – Each tenant can host an arbitrary number of projects with 1-N users assigned to them. A project is a high-level construct with the goal to create a unique product or service, such as a new “Door Lock” system software. The term project is used in a broad scope. Anything can be classified as a project.
  • Workbenches – A project consists of 1-N well-defined workbenches. A workbench represents a fully configured development environment of a specific “Workbench Type”. For example, a workbench of type “Simulation” allows for configuration and execution of Simulation Jobs based on drive data. Each workbench is implemented via a well-defined number of AWS Accounts, which is called an AWS Account Set.
    • An AWS Account Set always includes at least a Toolchain Account, Dev Account, QA Account and Prod Account. All AWS Accounts are baselined with IAM resources, security services and potentially workbench specific blueprints so development can start quickly for the end-user.

The following diagram illustrates the high-level architecture:

Figure 2 – High-level architecture diagram

Figure 2 – High-level architecture diagram

The CAEdge framework uses a data mesh architecture using AWS Lake Formation and Glue to create the tenant-level data lake. The Reference Architecture for Autonomous Driving Data Lake is used to design the Simulation workbench.

Implementation Details

With the core architecture and terminology defined, let’s look at the implementation details of the architecture that was described in the preceding image.

Isolated Tenants – Achieving a High Degree of Separation

To achieve a multi-tenant environment, we followed a multi-layered security hardening approach:

  • Tenant Separation on AWS Account Level: Starting at the AWS Account level, we used individual AWS Accounts where possible. An AWS account can never be assigned to more than one tenant. The functional scope of an AWS Account is kept as small as possible. This increases the number of total AWS Accounts, but significantly reduces the blast radius in case of any breach or incident. Just to give an example:
    • Each Dev, QA, and Prod Stage of a Workbench is its own AWS Account. No AWS Account ever combines multiple stages at once.
    • Each CAEdge tenant-owned data lake consists of multiple AWS Accounts. A data lake also requires updates as time passes. To allow for side-effect free and well tested updates of the data lake-infrastructure, each tenant comes with a Dev, QA, and Prod data lake.
  • Tenant Separation via a well-defined Organizational Unit (OU) structure and Service Control Policies (SCP): Each Tenant gets assigned a dedicated OU structure with multiple sub-OUs. This allows for tenant-specific security hardening on SCP-level and potential custom security hardening, in case dedicated tenants have specific security requirements. The SCPs are designed in such a way to allow for a maximum degree of freedom for the individual AWS Account user; while at the same time protecting the integrity of CAEdge and while staying compliant and secure according to specific requirements.
  • Tenant Separation through an AWS Account Metadata-Layer and automated IAM assignments: The CAedge framework uses a central Amazon DynamoDB database that maps AWS Accounts to Tenants and stores any other Metadata in the Context of an AWS Account. This includes including the Account Owner, Account Type, and Cost-related information. With this database, we can query AWS Accounts based on specific Tenants, Projects, and Workbenches. Furthermore, this forms the foundation of a fully automated permission and AWS Account access-management capability that enforces any Tenant, Project and Workbench boundary.
  • Tenant Separation Security Controls via AWS Security Services: On top of the standard AWS security services, such as AWS GuardDuty, AWS Config, AWS Inspector and AWS SecurityHub, we use IAM Access Analyzer in combination with our DynamoDB Account Metadata Store to detect the creation of any cross-account permissions that span outside of the AWS Organization, or that may have Cross-Tenant implications.

Automated creation and management of Tenant-Owned Resources and Services, Projects and Workbenches

CAEdge follows the “Everything-as-an-API Approach” and is designed as a fully open platform on the internet. All key features are exposed via a secured, public API. This includes the creation of Projects, Workbenches, and AWS Accounts including the management of access rights in a self-service manner for authorized users, as well as any updates affecting subsequent long-term management. This can only be achieved through a very high degree of automation.

We architect the following services to achieve this high degree of automation:

  • AWS Control Tower – An AWS managed service for account creation and OU assignment.
  • AWS Deployment Framework (AWS ADF) – an extensive and flexible framework to manage and deploy resources across multiple AWS Accounts and Regions within an AWS Organization. We use ADF to baseline all accounts with the resources required. This includes all security services, default IAM Roles, network related resources, such as VPCs and DNS and any other resources specific to the AWS Account type.
  • AWS Single Sign-On (AWS SSO) – A central IAM solution to control access to AWS Accounts. AWS SSO assignments are fully automated based on our defined access patterns using our custom Dispatch application and an extended version of the AWS SSO Enterprise solution.
  • AWS DynamoDB – A fully managed NoSQL database service for storing tenant, project and AWS Account data. Including information related to ownership, cost management, access management.
  • Dispatch CAEdge Web Application – A fully serverless web application that exposes functionality to end-users via API calls. It handles authentication, authorization, and provides business logic in the form of AWS Lambda functions to orchestrate all of the aforementioned services.

The following diagram provides a high-level overview of the automation mechanism at the platform level:

Figure 3 – High-level overview of the automation mechanism

Figure 3 – High-level overview of the automation mechanism

With this solution in place, Continental enables OEMs, suppliers, and other partners to spin up developer workbenches in a tenant context within minutes; thereby reducing the setup time from weeks to minutes using a self-service approach.


In this post, we showed how Continental built a secure multi-tenant platform that serves as the scalable foundation for software-intensive, vehicle-related workloads. For other organizations experiencing challenges when transforming into a software-centric organization, this solution eases the onboarding process so developers can start building within hours instead of months.

Martin Stamm

Martin Stamm

Martin Stamm is Principal Expert for Software Architecture at Continental Automotive. He has 32 years of professional experience in Software Engineering and has worked in the Automotive Software domain since 2003. After several years in Software Project- and Line Management, Martin turned his focus to platform and technology strategies. Since 2019, he has driven the transformation towards cloud-based and cloud-native solutions at Continental Automotive and is the technical lead of the CAEdge framework.

David Crescence

David Crescence

David Crescence is a Sr. Engagement Manager at AWS based in Munich, Germany. He leads AWS consultants, partners and client teams to deliver AWS products and services that help our customers realize their business objectives. He has extensive experience helping global customers in the Manufacturing and Automotive industry delivering innovative and cutting edge solutions on AWS. During his off-time David, loves to travel with his family back to La Réunion, an oversea French Island in the Indian Ocean.

Daniel Krumpholz

Daniel Krumpholz

Daniel Krumpholz is an Engagement Manager at AWS Professional Services. He leads multiple engagements with AWS customers and works with them to achieve their business outcomes. He also guides Professional Services teams on the delivery of the most complex engagements within AWS.

Andreas Falkenberg

Andreas Falkenberg

Andreas is Senior Consultant at AWS Professional Services. As a cloud infrastructure and security architect, he builds scalable, automated and secure platforms with his customers in the manufacturing vertical.

Junjie Tang

Junjie Tang

Junjie is Principal Consultant at AWS Professional Services. As a global technical lead, he heads a big data community of AWS Professional Services Consultants to develop data strategies and build accumulated expertise in data analytics across verticals.