IBM Support

Prerequisites and Maintenance for IBM Watson Machine Learning for z/OS

Release Notes


Hardware and software requirements for installing IBM Watson Machine Learning for z/OS on IBM z.


 IBM Watson Machine Learning for z/OS is an end-to-end enterprise machine learning platform that helps you create, train, and deploy machine learning models to extract value from your mission critical data on IBM Z, while keeping the data where it resides.

What's New


  • New ability to configure Jupyter notebook server for model training natively on z/OS using Spark and Python runtimes.
  • Introduction of ONNX model inferencing natively both in the Liberty scoring server on z/OS and through the LINK command for CICS® applications.
  • Enhanced performance of ONNX model inferencing by leveraging the advanced model conversion utility of IBM® deep learning compiler (DLC) and the on-chip AI accelerator of IBM Telum processor on z16.
  • New ability to micro-batch scoring requests of ONNX models in both the Liberty scoring server and the CICS-integrated scoring server.
  • New option for validating WML for z/OS models with Watson™ OpenScale on Cloud Pak for Data through the seamless integration of the two enterprise-grade machine learning and AI solutions.
  • New ability to customize z/OS jobnames for WMLz services and processes, which simplifies the management of the application tasks across your enterprise environment.
  • New ability to support in-transaction scoring capability through native WOLA interface for IMS COBOL applications. This native WOLA interface also applies to Batch COBOL applications, and CICS COBOL applications.
  • New ability to configure ONNX Compiler service on Linux® on Z, which lifts the requirement of having z/OS Container Extensions (zCX) for importing ONNX models.


  • Introduction of the WML for z/OS installation planner, a new Cloud-hosted tool with which you easily create a high level installation plan based your use case. To make your WMLz 2.3.0 planning easy, consider using the installation planner.
  • Integration of the optional Db2® anomaly detection solution into the WML for z/OS base, which significantly simplifies the installation and configuration of the solution.
  • Performance enhancements in online scoring of ONNX models with the addition of an MLIR-based compiler. The new compiler optimizes ONNX model inference and improves its deployment efficiency.
  • Improved deployment architecture that supports on-demand configuration of machine learning runtime environments. With this release, you install and configure a machine learning runtime environment only when you need it. For instance, you have the option of not configuring any Python runtime if your machine learning workload is SparkML only.
  • Improved up-and-running experience by removing node.js as a prerequisite and consolidating it as part of the WML for z/OS base installation.
  • New look-and-feel and enhanced performance of the WML for z/OS configuration tool. This release completes the transformation of the configuration tool UI based on the standards and principles of carbon, the IBM®'s open source design system for optimizing product interfaces and digital experiences.
  • Upgrade of the Scikit-learn library to version 0.23.
  • Introduction of the Online Scoring Community Edition (OSCE) of WML for z/OS. The OSCE is a special no-charge option for you to try out the WMLz feature of real-time scoring of pretrained ONNX models. You can download the OSCE Docker image and user's guide at IBM Web Membership (IWM) site.


  • Introduction of the new WML for z/OS IDE. With v2.2.1, the IDE is changed from the inside out. The IDE now runs on Red Hat® OCP. This allows the IDE to utilize fully the included core services and the underlying machine learning infrastructure and resources. The IDE user interface uses the same framework as that of Cloud Pak for Data. The new UI unifies the application experience for data scientists and system administrators as they manage system environments, services, users, data, machine learning assets, and deployment spaces.
  • New look-and-feel for WML for z/OS user interfaces and administration dashboard. With this release, the WMLz UIs start to adopt the standards and principles of carbon, IBM®'s open source design system for products and digital experiences. Carbon makes it easier for you to know exactly how a web interface will behave and where you are in a workflow. This enables you to spend less time learning the interfaces and more time using the services they provide.
  • Usability enhancements in online scoring of ONNX models. You can now import pretrained ONNX models directly into the WMLz base, without having first to convert them into a compressed format by using a specialized utility. Also, the ONNX model scoring service now runs natively in IBM z/OS Container Extensions (zCX), which makes the scoring workload zIIP eligible.
  • Performance enhancements in online scoring in general if your machine learning pipelines includes the following transformers:
  • Improved ease of creating data sets. You can now create new data sets by running SQL queries on a data connection in the integrated SPSS® flow. This allows your data scientists more flexibility to explore and analyze the data, instead of trying to access it.


  • New option of configuring a RACF® keyring-based keystore for user authentication.
  • New high availability support in WML for z/OS base through the specification of SHAREPORT and Sysplex distributor for load distribution and balancing.
  • Ability to import Spark, Scikit-learn, PMML, and XGBoost models trained on IBM Cloud Pak for Data.
  • Upgrade of Scikit-learn library to v0.22.x and XGBoost library to v0.90.
  • Enhancement in the creation and configuration of both standalone and clustered scoring services.
  • Enhancement in Db2 anomaly detection solution in terms of increased model analysis accuracy and decreased storage usage (due to reduced volume of baseline data).
  • Removal of the built-in Visual Model Builder and associated services, which reduces the footprint of WML for z/OS IDE.
  • Removal of z/OS LDAP as a prerequisite.

Prepare for the installation

For information concerning the preparation of the product installation see the following instructions

Maintenance Level

maintenance level
APAR for IBM Z IZODA V1.1 maintenance level
2022 Dec 
IzODA Spark 2.4.0 (FMID HSPK120) with PTF UI81887 applied
IzODA Anaconda 3.7.0 with PTF UI76587 &UI75844 applied (FMID HANA110)
IzODA Mainframe Data Service 1.1 (FMID HMDS120) with UI71323 applied
2022 Sep 
IzODA Spark 2.4.0 (FMID HSPK120) with PTF UI79625 applied
IzODA Anaconda 3.7.0 with PTF UI76587 &UI75844 applied (FMID HANA110)
IzODA Mainframe Data Service 1.1 (FMID HMDS120) with UI71323 applied
2022 Jan
2021 Oct
2021 Sep
2021 Jul 
IzODA Spark 2.3.0 or 2.4.0 (FMID HSPK120) with PTFs UI70807 & UI72095 applied
IzODA Anaconda 3.7.0 with PTF UI72334 applied (FMID HANA110)
IzODA Mainframe Data Service 1.1 (FMID HMDS120) with UI71323 applied
V2.2 PH25093
IzODA Spark 2.3.0 or 2.4.0 (FMID HSPK120) with PTFs UI69709 & UI69713 applied
IzODA Anaconda 3.6.0 or 3.7.0 with PTF UI69675 applied (FMID HANA110)
IzODA Mainframe Data Service 1.1 (FMID HMDS120) with UI71323 applied

WMLz Certification for IzODA SPARK maintenance level

WMLz version \ IzODA PTFs      
UI81887 (2022 Aug)
UI80875 (2022 May)
UI79625 (2022 Feb)
V2.3.0 Certified Certified Certified

Browser Support

WMLz for z/OS supports the standard view (or desktop version) of the following browsers

- Mozilla Firefox (v54 or higher) 

- Google Chrome (v60 or higher)

Make sure that you run WML for z/OS UI, administration dashboard, and configuration tool  in the supported browsers.

[{"Type":"MASTER","Line of Business":{"code":"LOB35","label":"Mainframe SW"},"Business Unit":{"code":"BU058","label":"IBM Infrastructure w\/TPS"},"Product":{"code":"SSCMZ4","label":"IBM Machine Learning for z\/OS"},"ARM Category":[],"Platform":[{"code":"PF035","label":"z\/OS"}],"Version":"All Versions"}]

Document Information

Modified date:
01 February 2023