At a glance
| Source connectors | More than 70 sources according to AWS, including S3, JDBC databases and AWS data stores |
| Destination connectors | S3 data lakes, Amazon Redshift, JDBC targets and open table formats |
| Batch or streaming | Batch ETL jobs and streaming jobs |
| Change data capture | Job bookmarks for incremental loads; zero-ETL integrations for some AWS sources |
| Transformations | Spark (Python or Scala), visual transforms in Glue Studio, DataBrew for no-code preparation |
| Orchestration | Triggers and workflows that chain crawlers and jobs |
| Deployment | Serverless on AWS only |
| Licensing | Pay as you go per DPU-hour, plus Data Catalog storage and requests |
AWS Glue suits teams standardised on AWS that want serverless Spark ETL, a shared Data Catalog for S3 data lakes, Athena and Redshift, and per-second billing with no clusters to manage. It is AWS only, and custom jobs are Spark or Python code, so it needs more engineering than a managed connector service.
Overview
AWS Glue is a serverless service to discover, prepare, move and integrate data from many sources for analytics, machine learning and applications. Crawlers infer schemas and register tables in the AWS Glue Data Catalog, which Amazon Athena, Amazon EMR and Amazon Redshift Spectrum can query. ETL jobs run on a serverless Apache Spark engine and can be authored as code or visually in AWS Glue Studio; streaming jobs, interactive sessions and job notebooks are also available. AWS says Glue can connect to more than 70 data sources.
Each job runs on a chosen AWS Glue version that fixes its Spark and Python versions. AWS Glue 6.0 uses Spark 4.1.1, Python 3.13 and Java 17 and adds Spark Declarative Pipelines and Iceberg format version 3; AWS Glue 5.1 (Spark 3.5.6) is the default for new jobs that do not specify a version. In the us-east-1 price list, Glue 6.0 jobs are priced lower per DPU-hour than earlier versions.
Key features
- Serverless Apache Spark ETL jobs billed per second in DPU-hours, including streaming jobs
- Crawlers that infer schemas and populate the AWS Glue Data Catalog
- Data Catalog shared with Amazon Athena, Amazon EMR and Amazon Redshift Spectrum
- AWS Glue Studio visual job editor that generates code
- Flex execution class at a lower DPU-hour rate for jobs that are not time sensitive
- AWS Glue Data Quality and sensitive data detection
- Native support for Apache Iceberg, Hudi and Delta Lake tables
Pros and cons
Pros
- Serverless: no clusters to provision or patch
- Data Catalog shared across AWS analytics services
- Per-second billing, a cheaper Flex class, and lower Glue 6.0 rates in us-east-1
- Built-in support for Iceberg, Hudi and Delta Lake
Cons
- AWS only, so it does not help with multi-cloud data movement
- Custom jobs are Spark or Python code, which needs more engineering than a managed connector service
- Several older Glue versions are end of life and Glue for Ray is closed to new customers, so check the version policy before you build
Pricing
AWS Glue is billed per second for the Data Processing Units (DPUs) your jobs, crawlers and sessions use. In US East (N. Virginia, us-east-1), from the AWS Price List published 6 November 2026 (USD): ETL jobs, crawlers and interactive sessions USD 0.44 per DPU-hour; Flex jobs USD 0.29 per DPU-hour; Glue 6.0 and later jobs and interactive sessions USD 0.308 per DPU-hour and Glue 6.0 Flex jobs USD 0.203 per DPU-hour; memory-optimised jobs USD 0.52 per M-DPU-hour (USD 0.364 on Glue 6.0). The Data Catalog includes the first 1 million objects stored and 1 million requests a month free, then USD 1 per 100,000 objects a month and USD 1 per million requests. Zero-ETL integrations have no integration fee; ingestion from applications is USD 1.50 per GB. Rates vary by region.
Pricing checked on the vendor's official page on 7 November 2026. Prices change; confirm before buying.
Who it is best for
Teams on AWS building S3 data lakes or loading Amazon Redshift that want serverless Spark ETL and a shared metadata catalog.
Who should look elsewhere
- Multi-cloud or SaaS-heavy ingestion where a managed connector service fits better: see Fivetran alternatives.
- Teams without Spark or Python skills who want point-and-click pipelines across clouds.
Verdict
AWS Glue is a practical default for data integration inside AWS: serverless Spark, a shared catalog and per-second billing, with Glue 6.0 bringing Spark 4 and lower us-east-1 rates. Its limits are scope and skills: it is AWS only, and real pipelines are code. Keep an eye on the version support policy, since AWS retired several older Glue versions in 2026.
Frequently asked questions
How much does AWS Glue cost?
Glue bills per second in DPU-hours. In us-east-1 on 7 November 2026, standard ETL jobs were USD 0.44 per DPU-hour, Flex jobs USD 0.29, and Glue 6.0 jobs USD 0.308. The first million Data Catalog objects and requests each month are free.
Which AWS Glue versions are still supported?
AWS lists Glue 0.9, 1.0 and 2.0 as end of life on 1 April 2026. Glue 3.0, 4.0, 5.0, 5.1 and 6.0 are listed, with 5.1 as the default and 6.0 the newest.
Is AWS Glue for Ray still available?
Only to existing customers. AWS closed Glue for Ray to new customers from 30 April 2026 and recommends Ray on Amazon EKS as the alternative.
Sources
- AWS Glue pricing
- AWS Price List: AWS Glue, us-east-1
- What is AWS Glue (Developer Guide)
- AWS Glue versions
- AWS Glue for Ray end of support
Checked November 2026.
How we research reviews and what star ratings mean: our editorial method.