What is serverless?

Summary

• Serverless is a cloud computing model where the platform monitors the resources a workload needs, automatically allocates as much compute as required, and scales back down when demand decreases, so you never manage infrastructure directly.
• Databricks serverless compute is a fully managed service: Databricks runs a secure fleet of compute on your behalf, with no clusters, quotas, or network plumbing for you to configure.
• Compute starts in seconds because Databricks keeps warm pools of instances ready; serverless SQL warehouses typically start in about 2 to 6 seconds.
• An intelligent autoscaler scales capacity up and down in graduated steps and reduces capacity when it is no longer needed, and you are billed only when compute is assigned to your workloads.
• Serverless compute powers SQL warehouses, notebooks, jobs, pipelines, and model serving, with Photon and autoscaling enabled automatically.

What is serverless?

Serverless is a cloud computing model in which a serverless platform monitors the cloud resources a workload needs, automatically allocates as much compute as required, and then scales the infrastructure back down when demand decreases. This removes the need for developers and administrators to provision, manage, or maintain infrastructure directly. On Databricks, this is delivered by serverless compute, a fully managed service that runs your workloads on a secure fleet of compute that Databricks operates on your behalf. See What is serverless computing? for the broader concept.

How Databricks serverless compute works

• No infrastructure to manage. Databricks manages a large, secure fleet of compute for you, so cloud administrators no longer adjust quotas, create and maintain network resources, or connect to billing sources.
• Instant, fast startup. Compute starts up in seconds rather than minutes because Databricks keeps "warm pools" of instances ready. Serverless SQL warehouses typically start in about 2 to 6 seconds.
• Intelligent autoscaling. An intelligent autoscaler scales workspace capacity in graduated steps to meet demand, so workspaces always have the right amount of capacity provisioned. Photon and autoscaling are enabled automatically.
• Scale to zero. The autoscaler reduces provisioned capacity after a few minutes when it predicts the capacity is no longer needed, and you are billed only when compute is assigned to a workload, not for the time to acquire and set up instances.

Which workloads run on Databricks serverless compute

• SQL warehouses provide instant, Databricks-managed compute with very fast startup for analytics and BI.
• Notebooks run interactive Python and SQL with automatic scaling and no infrastructure to configure.
• Jobs and workflows run without configuring or deploying infrastructure.
• Pipelines run Spark Declarative Pipelines (Lakeflow / Lakeflow Spark Declarative Pipelines) without configuring infrastructure.
• Model Serving provides a highly available, low-latency service that automatically scales to meet demand.

Why Databricks serverless compute

• Fast, simple, and reliable. With instant, on-demand compute, you do not wait for infrastructure resources or overprovision for usage spikes, and serverless pipelines eliminate cold starts and improve throughput.
• Fully managed, no overhead. Serverless removes the burden of capacity management, patching, upgrading, and performance optimization. Databricks updates workloads automatically and safely upgrades to the latest engine versions, so you always get the latest performance and security benefits.
• Built-in reliability. Serverless compute shields workloads from cloud outages with automatic instance-type failover and a warm pool of instances that buffers against availability shortages, and workloads are protected by multiple layers of security and designed to be enterprise-ready.
• Billed for work done. A simplified pricing model means one bill to track, and you pay only when compute is assigned to your workloads. See Serverless compute for notebooks, workflows, and Lakeflow Spark Declarative Pipelines is generally available.
• Improved productivity. Near-zero cluster startup latency and improved query concurrency mean every user can run workloads without help from infrastructure experts.

Getting started

• Read What is serverless compute? and the broader compute overview.
• Review the best practice to use serverless architectures for performance and efficiency.
• See Serverless compute for notebooks, workflows, and Lakeflow Spark Declarative Pipelines is generally available for how it works across workloads.

FAQs

What does serverless mean?

Serverless is a cloud model where the platform automatically provisions and scales compute to match a workload and scales it back down when demand falls, so you never manage servers, clusters, or infrastructure yourself.

How fast does Databricks serverless compute start?

Compute starts in seconds rather than minutes because Databricks keeps warm pools of instances ready. Serverless SQL warehouses typically start in about 2 to 6 seconds.

Do I manage any infrastructure with serverless?

No. Databricks manages a secure fleet of compute on your behalf, so you do not configure clusters, quotas, or network resources; capacity, patching, upgrades, and optimization are handled automatically.

What workloads can run on Databricks serverless compute?

SQL warehouses, notebooks, jobs and workflows, Spark Declarative Pipelines, and Model Serving, all with Photon and autoscaling enabled automatically.

How am I billed for serverless compute?

You are billed only when compute is assigned to your workloads, not for the time to acquire and set up instances, under a simplified pricing model with one bill to track.