The Pega-DecisionEngine agents support Decision Management operations.
The Pega 7 Platform keeps a local cache of scoring models. Scoring model updates are regularly retrieved from the adaptive data store. The model update frequency is implemented by periodically triggering the UpdateAdaptiveModels agent (Pega-DecisionEngine ruleset, PRPC:Administrators access group). The agent runs the pxUpdateModels activity to retrieve model updates.
By default, the agent is scheduled to run every 30 seconds. The agent only retrieves scoring models required for executing the strategy and the models that are different from those in the local cache. Configure model update frequency through the Services landing page.
Large scale simulations are enabled by performing strategy execution in batch across system nodes. The assignment, queuing and management of large scale simulations is governed by the ProcessBatchJob agent configuration. The agent is scheduled to run with a given regularity (in seconds) to trigger checking assignments in the [email protected] workbasket.
If there are assignments, they will be queued to create threads based on the thread configuration for each node. The status of the work item is updated as it progresses in this process and you can monitor the assignment by viewing the instances in the workbasket. How many threads can be run in a given node is something that you define in the Topology landing page. You need to have the ProcessBatchJob agent configured in your ruleset to make use of this functionality.
Proposition cache works on a single Pega 7 Platform node. When the Pega 7 Platform runs on multiple system nodes connected to the same database, Decision Management uses the system pulse to ensure the consistency of propositions across all nodes. The proposition cache is invalidated when a proposition is saved (triggered by adding or changing a proposition) or deleted.
Adding records that result in the proposition cache becoming invalid is done through two declare trigger rules that run the pyRefreshPropositions activity (pyPropositionSaved and pyPropositionRemoved in Data-pxStrategyResult).
If your installation consists of different Pega 7 Platform nodes connecting to the same database, enable the proposition cache synchronization mechanism by adding the PRPC:Administrators access group to the Pega-RULES: Core Engine Processing Agent data instance for every active node.
Adaptive Decision Manager can capture historical data for reporting purposes. The ADM Data Mart is implemented by periodically triggering the ADMSnapshot agent (Pega-DecisionEngine ruleset, PRPC:Administrators access group).
The agent runs the pzGetAllModelDetails activity. This activity captures the state of models, predictors and predictor binning in the ADM system at a particular point in time and writes that information to a table using the Data-Decision-ADM-ModelSnapshot and Data-Decision-ADM-PredictiveBinningSnapshot classes.
By default, the agent is scheduled to run every 120 seconds. The Data Mart settings in the Adaptive Decision Manager section of the Services landing page allow you to define how often the activity runs to capture the state of models and predictor binning.