In the very end primary node must not process primitives but only do:

• dispatching(using full metadata knowledge)
• final processing of the intermediate query parts results

In the middle of the migration towards better cloud support primary node must have less generic data in its single dbroot1.

The combination of PrimProc and ExeMgr will be a single process with separate threads(can be discussed but we need this to work gracefully using systemd) that behaves like PrimProc and ExeMgr,e.g. open separate network ports.

Luckily Bar made a framework for daemonizing services.