Precise Network Performance Prediction¶

This documentation covers the background, theory, and results for my research, Precise Network Performance Prediction, ( $PNP^2$ ), the aim of which is to provide design-time analysis of distributed cyber-physical systems’ (CPS) and their applications’ network performance. The type of network performance we will focus on predicting is the application buffer size requirements and the application network traffic buffering delay. These factors were chosen because they relate directly to how well the application traffic is serviced by the system and they directly affect resource utilization of the application on the system.

Because the buffering resources available to applications may be fixed by the system, application developers need to know at design time whether or not the application’s data production versus data consumption will overflow the application’s allotted buffer space. By analyzing these systems, developers can assess these conditions at design-time and try to mitigate any predicted overflows before deploying the applications.

Furthermore, buffering delay/latency is an important metric for determining application network performance because in many of these CPS, interactions and computation have deadlines that must be met, for instance in an attitude control system where the physical dynamics of the system determine the required reaction time of the software from sensing to calculation to actuation. Developers need guarantees that such reaction times can be met, and for distributed systems, the buffering latency and transmission latency for network data affects these reaction times.

Such design-time performance analysis and feedback is critical to the development of robust safety- or mission-critical applications.