V2.3 Release Notes

What’s New:

Manage:

• Spark-based Model Support: support for registering and managing Spark-based models

• HDFS Asset: integration of HDFS-persisted “assets” for model artifacts as well as for input/output data sets

• Snowflake integration: integration of ANSI-compliant SQL ”assets” for job execution for Snowflake and other common enterprise databases (e.g. SQL Server, DB2)

• SparkSQL Asset: integration of SparkSQL ”assets” to define input/output data sets for model execution (scoring, metrics jobs, etc.)

Monitor

• Spark Runtime Support: integration of Spark as an ”external runtime”, whereby ModelOp Center orchestrates and manages the execution of jobs within Spark, persisting all metadata and job execution information with each version of a model

• Starter Monitoring Templates: additional out-of-the-box “starter MLC’s” for Spark Execution for drift monitoring, concept drift monitoring, and backtesting

Govern:

• Model Code Scanning: support for integration with Veracode for model code scanning, including verification of passing scans within a Model Lifecycle

• Credentials Management: support for integration with Vault for credentials management for core services

• MLC Templates: simplified out-of-the-box MLC templates that can be used/customized

User Experience:

• Jobs page: updated Jobs page to allow orchestration across Spark and other model execution platforms, as well as an intuitive wizard to launch jobs

• Create snapshot: provided an intuitive wizard to create snapshots

• MLC Tagging: modified MLCs to be case insensitive for model tags

• MLC Details: included an additional description of each MLC, including inputs/outputs and tags used

• MLC UX: provided additional visibility into MLC processing, including current state and any incidents