3 New Features, Enhancements and Limitations in this Release

New Features

  • Installer Upgrades
    • SSO_TOKEN generation has now been automated and is no longer required as install configuration.
    • 1-off Patch support will now be made available for MMG 8.1.3.2.0 onwards.
    • mmg-python-library package is now available in wheel format along with tar.gz.
  • Utilities
    • New shell script utility to perform sync up of workspace schema with Data Model as an offline process.
    • Scheduler Batches can now be exported/imported using a new utility.
  • General Features
    • Access control introduced for folders at workspace-level in File Manager.
    • Object visibility control introduced for Model Pipeline Drafts and Datasets. They can now be restricted as private and shared with selected users and user-groups.
    • Log cleanups for better readability.
  • Conda
    • Performance improvement for package search.
  • Workspace
    • Workspace creation, which can take several minutes, has now been made as a background process. Workspace cards will now appear even for ongoing and failed creation processes.
    • Clean rollback has now been implemented for STSA/Simulations type workspaces.
  • Data Management
    • New module introduced that allows for managing Data Model updates, Data Population, and its schedules and history.
  • Model Pipelines
    • Model Review:
      • New screen to do peer-review for Drafts of Model Pipelines. Allows developers to review and received feedback on various aspects like Conceptual Soundness, Data Quality, Implementation Testing and Model Governance Assessment. These are admin-configurable and will allow for introduction of custom review types.
      • Reviewers can also Clone the Drafts and update/execute to do validations on their own copy.
    • Private Credential Injection: Allows individual users to optionally use their private credentials (without saving them in the system) for database connections. If not provided, connections will fall back to services accounts that is configured in Data Stores.
    • Upper limit for Python memory consumption can now be enforced for Model Pipeline executions.
    • Stage-level view is now available in Pipeline Canvas for enhanced navigation.
    • Executions can now also be done at stage-level.
    • Session Management:
      • Allows users to create, manage and choose the runtime sessions.
      • Runtime sessions are now synced between Pipeline Canvas and Notebook view.
    • Introduced better controls to restrict database connections and file accesses to workspace-scoped Data Stores and Folders.
    • Git Integration: Configured Git URLs now show as dropdowns for selection.
    • Tags can now optionally be provided during Model Publish.
    • Draft visibility can now also be restricted to private and shared with selected users and user-groups.
  • Technology Stack Changes
    • Support for Java 17 and 21
    • Support for OS: OEL9, OEL8 and Solaris
    • Support for Python 3.8-12
    • Additional support for Oracle DB 23 Client (apart from existing 19)
    • The Product now carries updated versions of the following:
      • OJet 18.1
      • Data Studio 25.3.2
      • PGX 25.3.0
      • All 3rd party libraries upgraded to latest stable versions for improved performance and security.
Known Issues and Limitations
  • Tables deletion sync-up between schemas is not supported during Workspace edits.
  • Error log table creation fails if the column data types are LONG, CLOB, BLOB, BFILE, and ADT during Workspace data population.
  • Unable to perform Dataset cache with Modin Library.
  • The PDF of the Model Report does not contain data in the output section.
  • The Notebook initialization fails if the logged in username is not in Uppercase.
  • Deployment of Models that are published from the Model Summary screen will not promote the associated dependencies such as Graphs, Parameter Sets, Datasets, and Models. However, the same would work fine, if Models are published from within Pipeline UI.
  • Python 3.12 does not support apache-flink fully right now, therefore, installation on Python 3.12 might give some error.
  • Oracle-pypgx-client version 25.3.x python package requires numpy >=2.0.2, while base mmg python functionality still rely on numpy 1.x so creating separate conda environment is recommended for using both as per the requirement (python-env-install.sh will install oracle-pypgx-client==25.3.0 and numpy==1.26.4 but some functionality might be affected due to requirement on numpy >=2.0.2).

Limitations from past releases that have been remediated in this Release

In Model Catalog, some techniques with train signature containing spaces used to fail, this has been fixed. It is still recommended to not use unnecessary spaces when giving the signatures during technique creation.