currently we are testing Pega 8.3 for our application which is a pure business rule engine, which means the application acts as a HTTP Service and consists mainly of activities, flows and data transforms. So no case type, no portal used or the like.
So the question here is which nodetype is the proper one in order to achieve the best performance. I assume the nodetypes WebUser and Search can be excluded since nobody will log in to Pega Designer Studio or any corresponding portal.
What I did is trying to test the performance without any classification and also test the performance while giving all nodes / pods the classification -Stream node. If my measurements are correct -Stream nodetype was way faster than the nonclassified nodes. But it could be coincidence of course and something else lead to this improvement, however I did not change anything else, so both tests run on the very same code basis and infrastructure, only the nodetypes where changed.
Do you have any suggestions here? I already checked most of the Pega community and Pega help sites regarding this topic, however its not clear to me which one is the best for a pure business rule engine which is highly dependent on a good performance.
If you are not using Pega decisioning stuff (e.g., CDH - Customer Decision Hub), try to start with NodeTypes of BackgroundProcessing/Stream. Stream (which runs Kafka embedded) is required for all background jobs (queue processors - which is the long term replacement of agents). If possible, run dedicated Stream Nodes. To be clear, the traditional agents should still work, but they should be running on 'BackgroundProcessing' nodes as well.