ROSELINE: Enabling Robust, Secure and Efficient Knowledge of Time Across the System Stack

Abstract:

Central to the operation of cyber-physical systems (CPS) is accurate and reliable knowledge of time, both for meaningfully sensing and controlling the physical world state and for correct, high-performance and energy-efficient orchestration of computing and communication operations. Emerging applications that seek to control agile physical processes or depend on precise knowledge of time to infer location and coordinate communication, make use of time with diverse semantics and dynamic quality requirements. These requirements need to be reflected in the system stack that runs across the layers from resonators that provide the physical time basis to software services and applications that make use of timing objects and estimates. Traditional hardware-OS and OS-application interfaces handle time in primitive ways by treating it primarily as a hardware reading that moves in essentially raw form to higher layers which have no run-time influence and little visibility. In the process, the timing information is compounded by increasing levels of uncertainty with no measures to account or correct for errors. ROSELINE seeks to address the resulting overdesign, inefficiencies, and fragility by rethinking how knowledge of time is handled across the system stack. Adopting an holistic notion of Quality of Time (QoT) that captures metrics such as resolution, accuracy, and stability, it explores ways for clock hardware, OS, network services, and applications to learn, maintain and exchange information about time, influence each other’s behaviors, and robustly adapt to dynamic QoT requirements and to benign and adversarial changes in operating conditions. The intellectual merits of the project flows from the following research activities: (i) Auto-tuning time service that exploits the diversity of available and tunable local, remote, and ambient time sources to provide required QoT while optimizing resource usage; (ii) Mechanisms for coping with physical- and software-based adversarial manipulations of QoT; (iii) Computational models for estimating and bounding QoT achievable under specific network conditions and available clock sources, and for run-time tuning of time service and clock sources; (iv) Analytic methods for relating application-level performance metrics with QoT requirements for guiding OS-level timing service and for application-level adaptation in response to varying QoT; (v) API and language- level constructs for richer expression of QoT requirements at the application-OS boundary, and corresponding time synchronization protocol and OS kernel scheduling mechanisms; (vi) Innovative, integrated clock sources with highly tunable QoT vs. power operating point based on programmable compensation circuits, multiple MEMS resonators, and rich hardware interfaces enabling bidirectional information exchange with software; and (vii) Embedded platforms with architectural support for time-related functions in sensing, actuation and networking subsystems. The research will be validated via applications involving time synchronization in smart grid, structural health monitoring, underwater sensing, and networked control. Applications that will accrue increased efficiency and robustness from the more tunable treatment of QoT abound across the entire spectrum of embedded, mobile and Internet-scale CPS, and include important applications in smart grid, factory automation, infrastructure monitoring, and data networks. Transforming the way time and its quality are managed in hardware, OS, network, and applications presents valuable opportunities to integrate research and education while promoting interdisciplinary learning at K-12 through postgraduate levels.