ReleaseBytes
databricks Databricks Blog ·

Databricks Builds Self-Serve Infrastructure Vending Machine for Field Engineering

bloginfradatabricksengineer
feature announcement

Databricks has developed the Field Engineering Vending Machine (FEVM), a Databricks App that provisions isolated, governed, and use-case-specific cloud resources on demand. This addresses infrastructure challenges faced by their growing field engineering organization of over 7,000 people, enabling faster development and agent-first framework adoption. The system uses Terraform for provisioning across AWS, Azure, and GCP, with transparency and audibility built-in.

  • →Introduction of the Field Engineering Vending Machine (FEVM)
  • →Use-Case-Based Provisioning with Agent Integration
  • →Databricks Field Engineering Grows, Demands New Infrastructure
  • →Automated Resource Management and Lifecycle Control
  • →Centralized Control and Resource Persistence
Features (2) ›
  • Introduction of the Field Engineering Vending Machine (FEVM)

    FEVM is a Databricks App designed to provide engineers with self-serve, isolated, and governed cloud resources on demand. It abstracts complex provisioning into a simple interface, allowing engineers to describe their needs and receive configured environments quickly.

  • Use-Case-Based Provisioning with Agent Integration

    FEVM enables provisioning based on specific use cases, such as building demos or reproducing issues, rather than generic workspace requests. It integrates with an agent-first framework, allowing users to request environments via natural language commands and leverage AI agents for complex workflows.

Enhancements (2) ›
  • Automated Resource Management and Lifecycle Control

    The system utilizes Terraform for provisioning and maintains a state database to track all resources, ownership, and expiration. Automated notifications are sent for provisioning, approaching expiration, and deletion, ensuring transparency and audibility throughout the resource lifecycle.

  • Centralized Control and Resource Persistence

    FEVM provides centralized control over global configurations and resource limits, managing shared resources like Unity Catalog. This ensures that platform limits are not exceeded and allows for resources like catalogs to persist independently of associated workspaces.

Notes (1) ›
  • Databricks Field Engineering Grows, Demands New Infrastructure

    Databricks' field engineering team has expanded to over 7,000 people, necessitating a move from small, manually maintained shared workspaces to a more scalable and manageable infrastructure model. The growth amplified challenges in workspace isolation, governance, and cost visibility.

Read the original announcement →

https://www.databricks.com/blog/provisioning-agentic-era-how-databricks-built-self-serve-infrastructure-vending-machine

Related releases