Skip to main content

Pillars & Core Principles

Building a successful SaaS application or data product on Databricks requires more than selecting a deployment model. These pillars represent the architectural foundations and operational principles that enable your solution to scale securely, operate efficiently, and deliver value to customers.

The Pillars​

PillarDescription
GovernanceUnity Catalog patterns for multi-tenant data isolation, access control, and compliance
Cost ManagementTagging strategies, budget controls, and usage attribution for customer billing and margin analysis
Scale & LimitsWorkspace, account, and cloud provider boundaries with design implications for capacity planning
Agentic AIProduction agents on Databricks with deployment-model-aware authorization, Genie, memory, and tracing
AutomationInfrastructure as code with Terraform and Databricks Asset Bundles for repeatable deployments
Customer and User OnboardingWorkflows for provisioning customers and users with group-based access management

Why These Pillars Matter​

Governance as a Foundation​

Multi-tenancy, data isolation, and access control are not optional considerations—they are architectural requirements. Strong governance enables you to serve multiple customers securely while maintaining compliance and auditability.

Cost Management as a Business Requirement​

Understanding per-customer costs is essential for pricing accuracy, margin analysis, and scaling profitably. Without proper attribution through tagging, you cannot measure the true economics of your product.

Scale & Limits as Design Constraints​

Every platform has boundaries. Understanding workspace limits, account quotas, and cloud provider constraints upfront prevents costly architectural rework as you scale from pilot to production.

Automation as an Operational Imperative​

Manual processes do not scale. Infrastructure as code, CI/CD pipelines, and automated provisioning are essential for operating efficiently at scale.

Onboarding as a Customer Experience​

How you provision customers and users directly impacts time-to-value. Group-based access management and automated workflows enable seamless onboarding without operational overhead.

Agentic AI as a Product Capability​

Customer-facing AI assistants require a distinct architecture: deployment-model-aware authorization (per-user for Customer Managed, per-tenant service principal for Partner Hosted), governed tool access through Genie, scoped memory, and production tracing. Treat agent design as a first-class pillar—not an add-on to your data platform.

Cross-Pillar Considerations​

These pillars do not operate in isolation. Effective architectures integrate them:

  • Governance + Cost Management — Tag enforcement policies ensure all compute resources have proper attribution tags
  • Scale + Automation — Infrastructure as code enables rapid provisioning while respecting limits and quotas
  • Governance + Onboarding — Group-based access management simplifies user provisioning while maintaining security
  • Cost + Onboarding — Customer-level tagging at provisioning time enables accurate billing from day one
  • Governance + Agentic AI — Deployment-model authorization and Unity Catalog enforce data access at query time; Genie attribution preserves governed outputs in the partner UI
  • Agentic AI + Cost Management — Unity AI Gateway and MLflow tracing support per-user model cost attribution

Getting Started​

Choose the pillar most relevant to your current architectural challenge:

What's Next​