Mainframe rehost + Airflow — modernize JCL and COBOL without a rewrite
Airflow doesn't run z/OS JCL. But once you rehost JCL and COBOL onto a distributed COBOL runtime, Airflow becomes the perfect orchestrator. BatchFoundry plans, executes, and operationalizes the full path.
Why Airflow alone isn't a mainframe replacement
Apache Airflow is an orchestrator — it schedules and monitors work, but it doesn't provide a runtime for the work itself. On z/OS, JCL job steps invoke COBOL or Assembler programs that depend on z/OS subsystems: JES2/JES3 for job entry and spool management, RACF for security, DB2 for z/OS, IMS, VSAM datasets with mainframe-specific access methods, and optionally CICS for transaction management inside batch. None of those subsystems exist in a pure Airflow cluster.
A "scheduler swap" — replacing Control-M or TWS with Airflow while keeping the COBOL on the mainframe — is a valid first step and BatchFoundry can execute that. But a full mainframe exit requires moving the runtime, not just the scheduler. That is the problem this solution page addresses.
The two-step modernization pattern
Rehost the runtime
- JCL → distributed JCL emulator (Raincode, Micro Focus)
- COBOL recompiled and run under a distributed COBOL runtime
- VSAM datasets migrated to PostgreSQL, Db2 LUW, or object storage with VSAM-compatible access layers
- DB2 for z/OS → Db2 LUW or Amazon RDS for PostgreSQL with schema conversion
Orchestrate with Airflow
- COBOL programs now run as Linux processes — Airflow schedules them with BashOperator or KubernetesPodOperator
- JES2/JES3 job nets become DAG edges with explicit dependency declarations
- Restart-from-step semantics map onto Airflow XCom + trigger_rule
- JES spool output → Airflow task logs and S3 remote logging
What BatchFoundry delivers
JCL inventory & dependency graph
We extract every JOB, STEP, DD, and PROC — building a full dependency graph before any code moves.
Target-runtime selection
Vendor-neutral cost and risk comparison across Micro Focus, Raincode, AWS MM, BMC, and Heirloom.
Airflow DAG generation from JCL job nets
Once COBOL runs on Linux, we generate Airflow DAGs from the JES2/JES3 job dependency net — the same way we handle any other scheduler.
Parallel-run cutover
Side-by-side execution against the live mainframe with output-equivalence checks before you flip the switch.
Supported target runtimes
Micro Focus Enterprise Server
Full COBOL + JCL emulation on Linux/Windows. Industry-standard for mainframe rehosting.
Raincode JCL/COBOL
Cloud-native JCL emulator and COBOL compiler. Strong AWS and Azure integration.
AWS Mainframe Modernization
Managed service offering both Blu Age (automated refactor) and M2 Runtime (rehost) paths.
BMC AMI Cloud
Lifts mainframe workloads into a managed cloud environment with JES and VSAM compatibility.
Heirloom Computing
Transpiles COBOL to Java. Suitable when a language-level modernization is acceptable.
What we won't tell you
If your batch depends on CICS transactions inside JCL EXEC PGM steps, or on VSAM datasets with massive concurrent update patterns, rehosting is hard. Some shops have IMS hierarchical databases or proprietary teleprocessing monitors that have no clean equivalent on distributed infrastructure. We will tell you up front — clearly and early — when refactor-to-Java, stay-on-mainframe, or a hybrid model is the better answer. We would rather lose a deal than set you up for a failed migration.
Schedule a 30-minute mainframe assessment
Tell us how many JCL job streams you run, your target environment (AWS, Azure, on-prem), and your programme timeline. We'll give you a straight answer on feasibility.
Schedule a 30-minute mainframe assessment