The state of the art in verifying the real-time requirements of applications developed using general processing graph models relies on simulation or off-line scheduling. We extend the state of the art by presenting analytical methods that support the analysis of cyclic processing graphs executed with on-line schedulers. We show that it is possible to compute the latency inherent in a processing graph independent of the hardware hosting the application. We also show how to compute the real-time execution rate of each node in the graph. Using the execution rate of each node and the time it takes per execution on a given processor, the resulting CPU utilization can be computed, as shown here for the Directed Low Frequency Analysis and Recording (DIFAR) acoustic signal processing application from the Airborne Low Frequency Sonar (ALFS) system of the SH-60B LAMPS MK III anti-submarine helicopter.