AWS Glue 6.0 cuts prices 30% with Iceberg v3 support

Amazon Web Services has launched AWS Glue 6.0, cutting pricing by 30% compared with earlier versions.

The update adds full support for Apache Iceberg v3, an open table format for managing large analytic datasets, and is available in all regions where AWS Glue operates.

AWS Glue is the company's serverless data integration service for extract, transform and load work, data discovery, and metadata management. Version 6.0 is built on Apache Spark 4.1, Python 3.13, and Scala 2.13, reflecting a broader refresh of the runtime customers use to prepare and process data.

One of the most notable changes is support for the Apache Iceberg v3 specification on Iceberg 1.11.0. This includes the VARIANT data type with shredding support, intended to improve read performance for semi-structured data compared with storing the same information in string columns.

The feature is aimed at organisations handling JSON records, logs, and event streams without flattening schemas before analysis. This approach can reduce duplicate data copies and avoid custom parsing code when source schemas change.

Glue 6.0 also adds support for Geometry and Geography data types, extending the service to spatial processing tasks such as geographic information system analysis and location data pipelines. Another addition is nanosecond-precision timestamps for workloads such as sensor processing, scientific data, and financial systems where sub-millisecond precision matters.

The update also includes unknown type handling, aimed at data pipelines that must continue running even when upstream schemas evolve unexpectedly.

Runtime refresh

Beyond Iceberg support, the release centres on Spark 4.1. One of the main additions is Spark declarative pipelines, which let engineers define the result they want from a transformation while the engine determines execution order and optimisation steps.

That shift reflects a broader move across data platforms to reduce manual orchestration in ETL design. For teams managing complex workflows, it means less hand-built dependency management and fewer steps when authoring jobs.

Another part of the update is Arrow-native execution for Python user-defined functions and user-defined table functions. By reducing data serialisation between Python and the Java virtual machine, the approach improves PySpark performance for more complex transformations.

Glue 6.0 also adds a real-time streaming mode for stateless streaming use cases. According to AWS, the mode can reach single-digit millisecond latency, making the service more relevant for workloads that require near-immediate event handling and low-latency routing.

Migration path

Customers do not need to make API changes to adopt the new version. They can choose Glue 6.0 through the existing version parameter when creating or updating jobs across the command-line interface, software development kits, Glue Studio, SageMaker Unified Studio, and integrated development environments.

Existing jobs can be upgraded through a Spark upgrade agent in Glue Studio or by using an automatic upgrade option. Notebook users can switch versions by setting Glue 6.0 in the session configuration for AWS Glue Studio notebooks and interactive Jupyter sessions.

The pricing cut may be one of the more commercially significant parts of the launch. As cloud customers continue to scrutinise data processing spend, lower pricing for a managed ETL service may matter as much as support for newer open table standards.

Competition in this segment has increasingly focused on interoperability with open-source formats such as Iceberg, Delta Lake, and Hudi, as customers try to avoid locking critical datasets into a single analytics engine or storage layer. By adding full Iceberg v3 support, AWS is responding to demand for tools that work with open data architectures while remaining fully managed.

The update also highlights how cloud providers are blending batch ETL, interactive analytics, and low-latency stream processing into fewer services. Glue began largely as a managed data preparation tool, but the latest release shows how providers now expect a single service to handle schema evolution, open table formats, Python-heavy engineering workflows, and real-time event processing.

AWS charges Glue users by the hour, billed by the second, for crawlers and ETL jobs. The Data Catalog carries a monthly fee for metadata storage and access, with the first million stored objects and the first million accesses available at no charge.