Pega performance affected due to ADM,RTDG and Realtime(inbound)

General
, , , , , , , ,
1

we have the following node configuration :

6 nodes for batch,backgroud ,custom1,custom2,search

4 nodes for RTDG,ADM and Realtime(Inbound)

1 search node

2 stream node.

1 webuser node.

We are using Pega CDH where were are running outbound and inbound campaigns where it process and loads heavy data .whenever there is some problem in batch run ,we also see some problem in Inbound where the avg response time spike is very high at that particular time.This does not happn every time but we have observed it some time .

We want to seprate and outbound process in a way when non of these are impacting each other when there are problem in any one outbound node, can you please review our node configuration and suggest any improvements?.

if there is any batch run ,will RTDG or ADM nodes are used for any purpose? if yes ,if any of the batch nodes gets stuck during batch run ,will it affect the RTDG or ADM process which eventually impact the inbound processing due to the above node configuration and how?

2

@Ashish Kumar I would recommend to create a support incident. Perhaps there is some contention for database/Cassandra resources but that’s just conjecture at this point,

Regards,

Ivar

3

@Ashish Kumar

When you open a ticket with Support, please mention this thread and also let us know the ID so we can track for you.

Thanks!

4

@Ashish Kumar I’ve just searched our pega support platform and saw that you had already logged INC-B1901 ( Pega performance affected due to ADM,RTDG and Realtime(inbound)) on 16th January.

We would appreciate it if in future you quote any open support tickets in conjunction with questions asked on this forum.

I see that despite the SME having asked you to request help from our support team, the ticket was closed and you were told to get help from your AE as this is a general product architecture question.

:warning: This is a GenAI-powered tool. All generated answers require validation against the provided references.

To optimize batch nodes for campaign batch runs, you can consider the following steps:

  1. Assign additional batch processing nodes if data job runs are hanging due to large volumes of data to process.
  2. Break down large work items into smaller ones and process them individually using a queue processor. You can scale queue processors horizontally by increasing the number of nodes configured with the node type, and vertically by increasing the number of threads.
  3. In the Data Flow service, specify the number of Pega Platform™ threads that you want to use for running data flows in batch mode. The higher the number of nodes for a mode, the higher the use of the mode.
  4. If all of the nodes in your environment are part of the DNode cluster, there is no separation. To achieve the separation, you need to set the thread count to 0 on the nodes that will be inbound designated nodes.

Remember, the configuration of your nodes and threads can significantly impact the performance of your campaign batch runs.

Node types for Pega Cloud services environments > Decision Strategy Manager

Seperation of responsibility in a Pega Marketing Cluster

Pega-DecisionEngine agents > Process batch job

Making decisions in data flow runs

Lack of Nodes

Tips for troubleshooting the Adaptive Decision Manager service

Node types for on-premises environments > Decision Strategy Manager node t

The ADM node fails to join the Cassandra cluster

@Ashish Kumar if you’re happy with the above answer, please mark this REply as Accepted Solution.

If you can consistently reproduce an actual performance problem, could you please re-open the ticket with clear a clear scenario for them to investigate.

@Ivar_Siccama can you provide some input here on the forum as the support team were not given an actual scenario to investigate?

5

@Ashish Kumar Real time data processing nodes are used in CDH for churning real time data in backend. These configs are all done in D_nodetype argument in SETENV.sh file.

For CDH this is required ideal is check the admin studio and see if the seperate nodes are assigned to RTDG and ADM, and not clubbed with background processing or web user nodes.