This video display is the forward view from an autonomous vehicle that has a frame rate /accuracy target to meet. We programmed the environment to emulate a series of unexpected fan and core failures, an example of the challenges of unpredictable workloads. 

The system has a variable configuration space, programmed to be Self-Aware. The frame rate goal(s) is indicated on the graph as is the companion graph of accuracy.

Without being Self-Aware, this system stopped functioning and the vehicle came to a complete stop, not able to process enough pixels to keep operating under the pre-programmed rule set. 

Being Self-Aware, with an accuracy goal, permitted this dynamic system to dynamically adjust the frame rate to maintain the accuracy target given the variable content of the environment being sensed. The vehicle was able to continue operating and rebounded when the fan failures and cores were re-established.

Notable is the demonstration of unpredictability, whether in the workload at a particular instant, the software configuration space or the hardware. No system is perfectly built and no workloads are perfectly predictable. Goals in a Self-Aware system can extract performance dynamically from the actual conditions of the moment, regardless of how the system or the workload is varied.