Use case · Cross-industry
Your data platform taken over, jobs and data included
A tool that changes owner, a Hadoop estate due for renewal, a SaaS contract to renegotiate: graal takes over your Python and Spark jobs, recreates your workflows and reads your data where it lives during the switchover. At the end, everything sits in open formats.
The challenge
Migrate without stopping production
- Dozens of jobs and workflows in production, written over the years.
- A Hadoop estate — HDFS, Hive, Kerberos — or a SaaS you need to leave.
- SAS programs still at the heart of some processing.
- Continuity of service expected throughout the switchover.
How graal handles it
A takeover in batches
Step 01
Inventory
graal Services list jobs, workflows, sources, permissions and dependencies with you, and sort out what moves as is, what needs adapting and what stops.
Step 02
Move the data
The Hadoop bridge reads Kerberized HDFS and Hive; tables migrate to Iceberg, on your S3 storage, at your pace.
Step 03
Move the processing
Python and Spark jobs are uploaded with their libraries, SAS programs and .sas7bdat files run on graal, and workflows are recreated.
Step 04
Switch over
Both platforms run in parallel, results are compared, then graal takes over, batch after batch.
The typical scenario: a data department that has been running a DataOps platform or a Hadoop cluster for several years, and has to decide what comes next. The 8-week pilot moves a first batch of jobs end to end, on your infrastructure, and prices the rest.
On 20/06/2025, Scaleway announced the acquisition of Saagie, a French DataOps platform, to build its data and AI platform.
From 12/01/2027, a provider of data processing services may no longer charge switching fees (Data Act, Article 29).
What you get
- Your Python and Spark jobs in production on your infrastructure
- Data in open formats: Parquet, Iceberg, SQL
- Workflows versioned in Git and documented
- A platform you can leave, when the day comes, with your code and your data
What stays with you
- The code of your jobs and pipelines
- Your data, as Parquet and Iceberg on your storage
- Your workflows and their Git history
Capabilities involved
- Connectors
Kerberized HDFS and Hive, databases, S3 and SAS files.
- Orchestration and compute
Python, Spark and SAS jobs, workflows.
- Lakehouse and SQL
Iceberg tables and federated SQL with Trino.
- Governance
Permissions and audit carried over project by project.
- Openness
Open formats, REST API and exportable code.
Frequently asked questions
Is graal an on-premises alternative to Databricks?
graal installs on your Kubernetes, on-premises or with the hosting provider of your choice, and your Spark and Python jobs move onto it. When your data must stay within your infrastructure, that is what graal is built for.
How long does a takeover take?
It depends on the estate. The 8-week pilot moves a first batch end to end and prices the rest.
Who carries out the migration?
graal Services, with your teams, or your teams alone, backed by the documentation and support.
Can we keep Hadoop during the transition?
Yes. The Hadoop bridge reads HDFS and Hive throughout the switchover; you shut down the old cluster once the results match.
Get your takeover priced
A first batch of jobs moved in 8 weeks, and the plan for the rest.