What are the advantages of going serverless?

Summary

• No infrastructure to manage. Databricks handles cluster configuration, capacity, patching, upgrades, and performance optimization, so you no longer pick instance types or tune Spark settings.
• Instant startup. Serverless compute starts in seconds from Databricks-managed warm pools; serverless SQL warehouses typically start in about 2 to 6 seconds.
• Automatic scaling and scale-to-zero. An intelligent autoscaler adds and removes capacity automatically and suspends inactive compute, so you pay only for what you use. Most customers see a 25% or greater reduction in serverless compute spend.
• Built-in resilience. Automatic multi-availability-zone failover, warm pools, and VM fallback shield workloads from zonal outages and cloud capacity shortages.
• Strong isolation and governance. Serverless compute is dedicated to you and never re-used, with unified Unity Catalog governance, automatic security updates, fine-grained access control, and customer-managed keys.

What are the advantages of going serverless?

Going serverless means Databricks manages the compute infrastructure for you, so your teams spend their time on data and AI work instead of provisioning and tuning clusters. Serverless compute is available across the Databricks Platform for SQL warehouses, jobs, notebooks, Lakeflow Declarative Pipelines, and Model Serving. The advantages fall into five areas: no infrastructure to manage, instant startup, automatic scaling, built-in resilience, and strong isolation and governance.

Why go serverless on the Databricks Platform

• No infrastructure to manage. With serverless compute you no longer choose instance types, configure cluster scaling parameters, or set Spark configurations. Databricks automatically manages capacity, patching, upgrades, and performance optimization, and cloud administrators are freed from managing quotas, network resources, and billing plumbing. See Use serverless architectures.
• Instant startup. Serverless compute starts in seconds because Databricks maintains warm pools of instances ready for immediate use. For serverless SQL warehouses, startup is typically about 2 to 6 seconds, which removes the wait for infrastructure to become available.
• Automatic scaling and scale-to-zero. An intelligent autoscaler scales workspace capacity up in graduated steps to meet demand and back down when demand falls, including scale-to-zero, which suspends inactive compute automatically with no manual intervention. Because you only pay for compute when it is assigned to a workload, most customers see a 25% or greater reduction in serverless compute spend. See What is serverless computing?.
• Built-in resilience. Serverless compute shields workloads from cloud outages with automatic instance-type failover and warm pools that buffer against capacity shortages. The infrastructure incorporates multi-availability-zone resiliency to withstand zonal outages, and VM fallback automatically switches to alternative virtual machines during capacity shortfalls.
• Strong isolation and governance. Serverless workloads run with multiple layers of isolation so compute is 100% dedicated to you and never re-used. You keep a unified governance model with consistent controls for data sharing and encryption, and Databricks applies security best practices automatically, including security health monitoring, automatic security updates, fine-grained access control, and customer-managed keys.
• Broad workload coverage. Serverless supports SQL warehouses, serverless jobs (with autoscaling and Photon enabled automatically), notebooks for interactive analysis, Lakeflow Declarative Pipelines for ETL, and Model Serving for deploying AI models with automatic scaling.

Getting started

• Read Use serverless architectures for the platform best practices.
• Turn on serverless SQL warehouses for interactive analytics and BI with near-instant startup.
• Move jobs, notebooks, and pipelines to serverless so workspace users can run them without cluster-creation permissions.
• Review What is serverless computing? to plan a migration path and estimate cost savings.

FAQs

Do I still manage infrastructure with serverless?

No. Databricks manages cluster configuration, capacity, patching, upgrades, and performance optimization. You choose the workload; Databricks provisions and scales the compute automatically.

How fast does serverless compute start?

Serverless compute starts in seconds using Databricks-managed warm pools. Serverless SQL warehouses typically start in about 2 to 6 seconds.

Does going serverless save money?

Yes. You pay only for compute assigned to a workload, and scale-to-zero suspends inactive compute automatically. Most customers see a 25% or greater reduction in serverless compute spend.

Which workloads can run serverless?

Serverless compute is available for SQL warehouses, jobs, notebooks, Lakeflow Declarative Pipelines, and Model Serving.