**2.2.2 Session startup and process spawning**

Considering the question which process should work on which subtask leads to a further question: When shall the processes of a session be created? Regarding this problem, two approaches can be distinguished: In the case of a *static* startup, all the processes are created at the beginning of a parallel run and are normally bound to their respective processors during runtime. Such a static startup is usually conducted and supported by a job scheduler detecting and assigning idle processors. However, in the case of a *dynamic* process startup, further processes can be spawned by already running processes even at runtime. This approach is commonly combined with the MPMD paradigm. So for example, when a master process running a master program spawns worker processes running subtask-related subprograms. However, this approach demands for an additional interaction between spawning processes and the runtime environment during execution, in order to place spawned processes onto free processors. That is the reason why this approach is more complicated in most cases.
