Published Release Notes

Create resilient data flow runs

The resilience of data flow runs has been improved. If Data Flow service nodes fail, active data flow runs continue on the remaining functional nodes. Depending on the type of primary data source and the selected run failure policy, data partitions on the failed nodes can be reprocessed on another functional node or the data flow run is marked as failed for further investigation. You can also add or remove nodes from the Data Flow service while data flow runs are in progress. The runs are automatically rescaled.

For more information, see Data flow run updates.

Define partitioning of stream data sets

You can define partition keys for Data Set rules of type Stream to group similar properties and to process Data Flow rules that contain stream data sets more efficiently across multiple Data flow service nodes. Use this feature only for testing in non-production application environments.

Create specialized Decision Data rules for text analytics resources

You can store various types of resources for text analytics in Decision Data rules that are based on templates. Each resource type (for example, sentiment analysis models, taxonomies, entity extraction rules, and so on) has its own definition class that defines the appearance of the corresponding Decision Data rule form. With this enhancement, system architects can create and maintain various text analytics-related artifacts without having to edit the default Decision Data form fields first.

For more information, see Definition class of text analytics Decision Data rules.

Unable to create text analytics models when Java 2 Security is enabled

Security exceptions that prevent you from creating text analytics models are caused by the Java 2 Security feature that is enabled at the JVM level. This feature denies access to the text analytics resources that are required for text parsing functions.

Creating a text analytics model results in a failure because of a number of security-related exceptions, for example:

java.security.AccessControlException: Access denied ("java.lang.RuntimePermission""createSecurityManager")

The suggested approach for avoiding this problem is to use the text analytics models that are provided by default, for example, pySentimentModels, pyTelecomTaxonomy, and so on.

For more information, see Text Analyzer.

Expose properties in Decision Data Store

You can now expose properties in Cassandra-based Decision Data Store data sets. With this solution, you can update single properties in extended customer analytic records (XCAR) without having to update the full customer record.

For more information, see XCAR in Decision Data Store.

Enhancements to bulk editing options for unversioned propositions

Bulk editing options for unversioned propositions have been enhanced to allow the use of Excel for editing. You can export and import unversioned propositions into Excel by using the Import wizard. Additionally, data types replace decision tables as the new storage format for unversioned propositions.

For more information, see Data type management and Proposition management enhancements.

Reuse existing propositions across applications

You can reuse existing propositions from built-on applications by using a new wizard to copy proposition groups across applications. This wizard gives you more options and flexibility when you define the business hierarchy in your application.

For more information, see Proposition management enhancements.

Migrate DDS data across various schema versions

You can use multiple versions of a single Decision Data Store (DDS) data set type in different applications that run on the same Pega^® Platform instance and efficiently move data across those versions even if the underlying data schema is different. This solution improves the data management process and simplifies the reuse of existing data.

For more information, see Decision Data Store migration.

Support for PMML version 4.3 in predictive models

You can now leverage predictive models in the Predictive Model Markup Language (PMML) version 4.3 by importing them into predictive models on Pega^® Platform. Pega Platform utilizes a large part of the PMML standard with the exception of some PMML 4.3 features that are not supported in the Predictive Model rule.

For more information, see Support for PMML version 4.3 in predictive models and Creating a predictive model.

Emit remaining event strategy data at the end of a data flow run

Data flows have been updated to prevent data loss in event strategies. By default, when a batch or a real-time data flow stops and it contains an event strategy with Tumbling windows, all of the events that are in the Tumbling windows are emitted. Event emission triggers further processing in the event strategy, like aggregations and joins.

For more information, see Enhancements to data flows that contain event strategies and Event Strategy rule form - Completing the Event Strategy tab.