Stated briefly, a job flows from creative application to processing engine as follows. The creative application—Smoke, 3ds Max, Lustre, etc.—submits the job to the Backburner Manager, which is then distributed as blocks of tasks to the Backburner Servers with the corresponding adapter installed, residing on the render nodes. The servers invoke the appropriate processing engine, which carries out its assigned tasks.
While it is the creative application that specifies the number of tasks in a particular job, once the job is on the Backburner network, it becomes the responsibility of the Backburner Manager. For example, it is the manager that decides how many tasks each render node is asked to carry out, in what order tasks are sent out, etc.
This is illustrated in the following diagram.
The process just described is largely automated and can operate entirely without user intervention. That being said, once a job is on the Backburner network, you can intervene in a number of ways. Aside from passively monitoring jobs, the most common job activity is to suspend a job, change its settings, and restart the job.