| Computing systems now monitor, coordinate, control, integrate and facilitate many physical processes from vehicle management and crisis response to managing data centers. Such systems, termed Cyber-Physical Systems (CPS), can consist of three major components---(i) human inhabitants, (ii) physical environment and (iii) computing entities. Trustworthiness of the CPSs depends on how safe the physical environment is and how survivable the human inhabitants are in the environment. Safety and survivability require pro-active operations in the CPSs so that the conditions violating these properties are predicted and avoided. Pro-activity, however, generally causes undesirable resource consumption overhead. Further, the complex interactions among the physical environment and computing entities can cause additional overhead and uncertainty in pro-actively ensuring the safety and survivability. Thus, a synergistic design of pro-activity is required which considers such complex interactions.;Further, to avoid redline temperatures for equipment safety in a data center, which is another example of CPSs, cooling systems are pro-actively pre-set for worst-case thermal conditions; thus wasting cooling energy. This dissertation develops a set of energy-efficient data center job scheduling algorithms that consider the cooling behavior, computing equipment power characteristics, and its impact on the cooling demand to minimize the data center energy consumption.;Lastly, pro-active routing protocols in Mobile Ad hoc NETworks (MANETs), the most common computing infrastructure for crisis response, maintain routes between any two nodes irrespective of data to transmit. This dissertation introduces autonomic tuning of the route update frequencies in such protocols to minimize the energy-overhead while maintaining the service reliability for information exchange. Further, self-managing routing protocols are developed to pro-actively construct energy-efficient routes in MANETs.;In this regard, synergistic planning and preparedness of crisis response, which is cyber-physical in nature, is performed to pro-actively avoid life losses during crises. To this effect, crisis response is modeled as a state-based, real-time stochastic system capturing the uncertainties due to human interactions. The research outcomes include a crisis preparedness tool using the industry standard Architecture Analysis and Design Language (AADL) to specify and analyze the proposed stochastic model. |
