Lifetime-Driven Scheduling of Security-Critical Internet-of-Things Applications on Real-Time Heterogeneous Multicore Systems

With the rapid advance in information technology, we have witnessed the emergence of diverse promising Internet-of-things (IoT) applications, such as smart transportation, smart home, healthcare, and environmental monitoring. Due to parallel computing advantages, multicore systems are advocated as a good candidate hardware platform to accelerate the computation completion of IoT applications. However, the limited energy budgets and inherent openness characteristics of IoT scenarios pose fundamental challenges to the design of multicore systems. While making salutary attempts on improving the energy efficiency and augmenting the security level of IoT applications, most of the existing research works neglect the negative impacts of their optimization techniques on the lifetime of multicore platforms. Consequently, multicore systems potentially suffer from the earlier occurrence of permanent faults, i.e., have to be replaced in advance before the desired lifecycle ending point. In this paper, we conduct the first study on scheduling precedence constrained IoT applications depicted by directed acyclic graph (DAG) to minimize the energy consumption of multicore platforms under timing, security, and lifetime constraints. To address this problem, we first present a JADE-based offline algorithm to determine the optimal task-to-core allocation, task execution order, core operating frequency, as well as task security level. Following the offline algorithm, we then design a low-complexity dynamic algorithm to achieve extra energy savings by wisely adjusting core operating frequencies. Results reveal that our solution can reduce 41.48% energy consumption while meeting all design constraints compared to state-of-the-art schemes.