Platform Engineering
Platform Engineering at AVS
Innovation through Standardized Development Environments
AVS has revolutionized software development with an innovative Platform Engineering approach, tackling challenges like standardization, governance, and compliance. At its core is a tailored Internal Developer Platform (IDP), ensuring consistent environments and intelligent abstraction between developers and engineers. Automation boosts efficiency and streamlines cloud-native application orchestration. Built on Humanitec, Score, Azure, AWS, and OpenShift, with GitLab and ArgoCD for seamless DevOps integration, the platform is treated as a product—prioritizing usability and rapid implementation. A Minimum Viable Platform (MVP) enables fast deployment and continuous improvement by expert teams.
reduced onboarding times
higher developer productivity
improved team satisfaction
stronger brand presence
Successful Customer Adoption
reduced onboarding times
higher developer productivity
improved team satisfaction
stronger brand presence
Successful Customer Adoption
Bechtle AVS empowers companies with this platform strategy to optimize their development processes and drive their digital transformation forward.
What were the challenges
Before implementing Platform Engineering, Bechtle AVS faced several operational and technical challenges. The lack of standardization led to inconsistent development environments and difficulties in maintenance. Compliance and governance were complex due to varying regulations and security requirements. Additionally, there was a high cognitive load on developers due to the diverse software needs and staff turnover. The public sector demands further complicated matters, with differing security and compliance expectations. To address these issues, Bechtle AVS adopted Platform Engineering to create a secure, standardized platform, streamline development processes, and ensure consistent application delivery, ultimately paving the way for a more efficient and modern workflow.
Technical core components used by AVS
The Bechtle AVS Internal Developer Platform (IDP) is a modern, integrated environment designed to optimize software development and deployment, especially in regulated settings. It focuses on automation, standardization, and security. Key components include Score for workload specification, Humanitec for orchestration, Git & GitLab for version control, and Backstage as the central developer portal. It uses Kubernetes, OpenShift, and Helm for container orchestration, along with Terraform and Ansible for automation. The platform supports Microsoft Azure, AWS, and Dell Apex Cloud, and uses Prometheus and Grafana for monitoring. Security is managed with Microsoft Entra ID and Keycloak, while data is handled by MariaDB, Minio, and Apache Kafka.
What is aInternal Developer Platform
An internal developer platform (IDP) is the set of all technologies and tools that a platform engineering team brings together to create an internal product (platform as a product) in order to accelerate and optimize software development within an organization.
Build your Mininmal Viable Platform (MVP)
A minimal viable platform (MVP) for an IDP focuses on delivering the core functionalities developers need most urgently. This could include:
self-service access to version control
automated build and deployment pipelines
basic infrastructure provisioning
By starting small and gathering feedback, the IDP team can iterate and expand the platform’s features based on what proves most valuable to developers. This lean approach reduces risk, saves time, and ensures the IDP solves the right problems from the beginning.
Building Blocks of an IDP
A reference architecture is a pre-built blueprint that helps organizations design and implement a system following best practices. In the context of internal developer platforms (IDPs), a reference architecture provides a common structure with key components to ensure consistency, simplify development, and promote automation for developers. These blueprints are often adaptable to different cloud environments and can be customized to integrate with existing tools.
Developer Control Plane
Developer Control Plane
- This is the primary interface for developers interacting with the platform
- Here, they can define workloads (like deployments) and access various developer tools offered by the IDP
Integration and Delivery Plane
Integration and Delivery Plane
- This plane bridges the gap between developers and the platform's operational aspects
- It handles processes like building application images, configuration management, and deployment based on developer input from the Control Plane
Monitoring and Logging Plane
Monitoring and Logging Plane
- This plane provides real-time insights into the health and performance of both applications and the underlying infrastructure
- It collects metrics and logs that can be crucial for troubleshooting and optimization.
Security Plane
Security Plane
- This plane is responsible for securing sensitive information within the IDP
- It manages secrets, API keys, and credentials to ensure restricted access and prevent unauthorized activity
Resource Plane
Resource Plane
- This plane represents the underlying infrastructure that runs the applications
- It encompasses resources like compute clusters, databases, storage, and DNS services
Inside the AI-Enhanced IDP Architecture
The AI-empowered Internal Developer Platform (IDP) architecture from Bechtle Competence Center AV Software Solutions 360° (AVS), developed in collaboration with NVIDIA and Dell, is designed to support machine learning (ML) development and deployment, particularly in industries with strict security and compliance requirements. AVS’s new architecture advances traditional platform engineering by integrating AI-centric features across three distinct planes:
AI-enhanced Developer Platform
User/Developer Plane
Specifically designed for ML researchers, data scientists, and AI Ops teams, this layer provides a highly intuitive, developer-friendly environment that includes:
- An integrated development environment (IDE) and user portal.
- Version control systems, service applications, ML pipelines, and feature stores.
- Support for inference workload management alongside BI/data science tools and model training capabilities.
- Seamless integration with industry-standard tools such as PyTorch, CUDA, SparkML, and MXNet.
Control Plane
Ensuring the platform's security, stability, and cost-effectiveness, this plane handles:
- Comprehensive monitoring and logging with advanced observability via Dash0.
- Rigorous security measures, including identity and secrets management along with strict policy controls.
- A cost management system that oversees reporting, billing, and chargebacks.
Resource Delivery Plane
This layer is all about execution-deploying and managing AI workloads efficiently:
- A robust continuous integration (CI) pipeline flow.
- A dedicated model registry and an orchestration engine.
- Deployment capabilities powered by accelerated computing infrastructure featuring GPUs and RDMA networks.
