databricks Platform · Updated Jul 23, 2026

Databricks Runtime 19

General Availability of Databricks Runtime 19 with enhanced performance features.

What it does

Databricks Runtime 19 introduces Arrow-optimized Python UDFs enabled by default, which enhances serialization performance. It also includes TopN pushdown for fine-grained access control, allowing `ORDER BY ... LIMIT` operations to be executed on the remote cluster, thus improving query performance. Additionally, the `parquet-java` library has been upgraded to version 1.17.0, which includes various enhancements and bug fixes.

Why it matters

These updates significantly improve the performance and efficiency of data processing tasks, particularly for organizations that rely on Python UDFs and require fine-grained access control. The enhancements can lead to reduced query execution times and better resource utilization, ultimately lowering operational costs and improving productivity for data teams.

Key capabilities

Arrow-Optimized Python UDFs

Regular Python UDFs now use Arrow-based serialization by default for better performance.

TopN Pushdown

Supports pushing down `ORDER BY ... LIMIT` operations to improve query performance on dedicated compute.

Parquet Library Upgrade

Upgraded `parquet-java` to version 1.17.0, enhancing functionality and fixing bugs.

Recommended use cases

Data processing tasks requiring high-performance Python UDFs.Queries needing fine-grained access control with efficient execution.Organizations leveraging Parquet for data storage and processing.

Limitations · remember

!Changing the default Arrow serialization may alter type-coercion behavior for some UDFs.

Architecture

Databricks Runtime → Python UDFs → Query Execution Engine → Data Access Layer

In short

Databricks Runtime 19 is now available, featuring improved Python UDF performance and enhanced query capabilities.

Read the full documentation ↗

Key takeaway

Databricks Runtime 19 enhances performance and efficiency for data processing with new features and optimizations.

2026-07-23-databricks-runtime-19-july-23-2026 Generated by the Databricks Updates Agent · browse all updates