Loading...

Thermal-Aware standby-sparing technique on heterogeneous real-time embedded systems

Ansari, M ; Sharif University of Technology | 2022

97 Viewed
  1. Type of Document: Article
  2. DOI: 10.1109/TETC.2021.3120084
  3. Publisher: IEEE Computer Society , 2022
  4. Abstract:
  5. Low power consumption, real-time computing, and high reliability are three key requirements/design objectives of real-time embedded systems. The standby-sparing technique can improve system reliability while it might increase the temperature of the system beyond safe limits. In this paper, we propose a thermal-aware standby-sparing (TASS) technique that aims at maximizing the Quality of Service (QoS) of soft real-time tasks, which is defined as a function of the finishing time of running tasks. The proposed technique tolerates permanent and transient faults for multicore real-time embedded systems while meeting the Thermal Safe Power (TSP) as the core-level power constraint, which avoids thermal emergencies in on-chip systems. Executing the main and backup tasks on the cores at any power consumption below TSP guarantees that no thermal violation occurs. Our TASS proposed method provides an opportunity to remove the overlaps of the execution of main and backup tasks to prevent extra power consumption due to applying the fault-tolerant technique. Meanwhile, in order to maximize the QoS, we employ a heterogeneous platform to execute the main tasks as soon as possible on high-performance cores with more power budget. The backup tasks are executed on low power cores after finishing the main tasks. In this case, when the main task finishes successfully, the whole of its corresponding backup task can be dropped, resulting in a significant amount of power and temperature reduction. Therefore, in the fault-free scenarios, the spare cores can be powered down, and only the main tasks are scheduled and executed on the primary cores. Experiments show that our proposed method improves QoS up to 39.78% (on average by 18.40%) and reduces the peak power consumption and temperature by up to 40.21% and 15.47°C (on average 28.31% and 13.60°C), respectively, at runtime, while keeping the system reliability at the required level. © 2013 IEEE
  6. Keywords:
  7. Power consumption ; QoS ; Real-time embedded systems ; Standby-sparing ; Thermal management ; Thermal safe power ; Budget control ; Computing power ; Embedded systems ; Green computing ; Interactive computer systems ; Power management ; Quality of service ; Real time systems ; Reliability analysis ; Temperature control ; Embedded-system ; Multi-core processing ; Power ; Power demands ; Quality-of-service ; Real - Time system ; Real-time embedded systems ; Standby-sparing ; Task analysis ; Thermal ; Thermal safe power ; Electric power utilization
  8. Source: IEEE Transactions on Emerging Topics in Computing ; Volume 10, Issue 4 , 2022 , Pages 1883-1897 ; 21686750 (ISSN)
  9. URL: https://ieeexplore.ieee.org/abstract/document/9582741