Holistic performance control for mission-critical cyber-physical systems

Recent years have seen the growing deployments of Cyber-Physical Systems (CPSs) in many mission-critical applications such as security, civil infrastructure, and transportation. These applications often impose stringent performance requirements on system sensing fidelity, timeliness, energy efficiency and reliability . However, existing approaches treat these concerns in isolation and hence are not suitable for CPSs where the system performances are dependent of each other because of the tight integration of computational and physical processes. In this dissertation, we investigate the dependencies between these performances and propose the holistic performance control approaches for two typical mission-critical CPSs, which are Wireless Cyber-phyiscal Surveillance (WCS) systems and data centers. We first propose a holistic approach called {em Fidelity-Aware Utilization Controller} (FAUC) for WCS systems that combine low-end sensors with cameras for large-scale ad hoc surveillance in unplanned environments. By integrating data fusion with feedback control, FAUC enforces a CPU utilization upper bound to ensure the system's real-time schedulability under dynamic CPU workloads at runtime because of stochastic detection results. At the same time, FAUC optimizes system fidelity and adjusts the control objective of CPU utilization adaptively in the presence of variations of target/noise characteristics. The testbed experiments and the trace-driven simulations show that FAUC can achieve robust fidelity and real-time guarantees in dynamic environments.;We then present a proactive thermal and energy control approach for data centers to improve the energy efficiency while ensuring the data center reliability. It consists of a high-fidelity real-time temperature prediction system and a predictive thermal and energy control (PTEC) system. The prediction system integrates Computational Fluid Dynamics (CFD) modeling, in situ wireless sensing and real-time data-driven prediction. To ensure the forecasting fidelity, we leverage the realistic physical thermodynamic models of CFD to generate transient temperature distribution and calibrate it using sensor feedback. Both simulated temperature distribution and sensor measurements are then used to train a real-time prediction algorithm. Based on the temperature prediction system, we propose the PTEC system, which leverages the server built-in sensors and monitoring utilities, as well as a network of wireless sensors to monitor the thermal and power conditions of a data center. It predicts the server inlet temperatures in real-time, and optimizes temperature setpoints and cold air supply rates of cooling systems, as well as the speeds of server internal fans, to minimize their overall energy consumption. To ensure the data center reliability, PTEC enforces a set of thermal safety requirements including the upper bounds on server inlet temperatures and their variations, to prevent server overheating and reduce server hardware failure rate. A partition-based approach is proposed to solve the control problem efficiently for large-scale data centers. Extensive testbed experiments and trace-driven CFD simulations show that PTEC can safely reduce substantial cooling and circulation energy consumption compared with traditional approaches, and can adapt to the realistic and dynamic data center workload.