Overview - Databricks Well-Architected Framework

Databricks Well-Architected Framework

The Databricks Well-Architected Framework provides guidance and best practices for designing, deploying, and operating intelligent data products on the Databricks platform. This framework helps you make informed architectural decisions based on your specific requirements around data residency, management overhead, intellectual property protection, and scalability needs.

Deployment Types

When building on Databricks, you can choose from several deployment models, each with distinct characteristics and trade-offs:

SaaS Deployment

A fully managed, multi-tenant deployment where Databricks manages the infrastructure and platform operations. You can deploy either as workspace-per-customer or multi-tenant workspace configurations.

Best for: Organizations prioritizing rapid deployment, scalability, and minimal operational overhead.

Managed Hosted Deployment

A deployment model where the control plane is managed by Databricks, but the data plane resides in your cloud account. This model provides greater control over data residency while maintaining managed platform services.

Best for: Organizations with specific data residency, compliance, or governance requirements, commonly used in government and regulated industries.

Shipped Deployment

A deployment where customers manage their own Databricks instance, with full control over both control plane and data plane. Requires a Databricks license to operate.

Best for: Organizations requiring maximum control, operating in air-gapped environments, or with strict data sovereignty requirements.

Choosing the Right Deployment Model

The choice of deployment model depends on several key factors:

Factor	SaaS	Managed Hosted	Shipped
Data Residency	Limited control	Full control	Full control
Level of Management	Fully managed by Databricks	Shared management	Customer managed
Safeguard IP	Shared environment	Isolated environment	Fully isolated
Scale/Velocity	High scalability, rapid deployment	Moderate scalability	High scalability with operational overhead

Key Considerations

Data Residency

If your organization has strict requirements about where data must reside (specific regions, countries, or cloud accounts), Managed Hosted or Shipped deployments provide the necessary control.

Level of Management

Consider your team’s capacity and expertise in managing infrastructure:

SaaS: Minimal operational burden, focus on data products
Managed Hosted: Moderate operational requirements
Shipped: Full operational responsibility

Protecting Intellectual Property

For organizations building proprietary data products or algorithms:

SaaS: Appropriate for most use cases with proper access controls
Managed Hosted: Better isolation with dedicated resources
Shipped: Maximum isolation and control

Velocity and Scale

Consider your growth trajectory and time-to-market requirements:

SaaS: Fastest time to value with elastic scaling
Managed Hosted: Balanced approach with some operational overhead
Shipped: Requires upfront infrastructure setup

Framework Structure

This framework provides detailed guidance for each deployment model across key architectural pillars:

Workspace & Multi-Tenancy: Strategies for organizing workspaces and supporting multiple customers
Product Design: Best practices for creating intuitive, persona-driven products
Data Isolation: Approaches to data segregation and security
Hub & Spoke Model: Patterns for distributing data across environments
Security & Authentication: Identity management and access control
Automation: Infrastructure as code and deployment automation
Cost Management: Chargeback, tagging, and cost optimization
Scale & Limits: Understanding platform limits and scaling strategies

Next Steps

Choose your deployment model to explore detailed architectural guidance: