Energy and Temperature-Aware Resilient Task Scheduling for Renewable Powered Real-Time Systems

image

Title: Energy and Temperature-Aware Resilient Task Scheduling for Renewable Powered Real-Time Systems
Directors: Anne BENOIT & Mingsong CHEN
Discipline: Computer Sciences
Status: Completed Project
Starting date: 2017

Directors

Summary

The past several years have seen an increasing interest in emerging applications of embedded systems such as wireless sensor nodes and biomedical implants, which are scavenging energy from their environment. Due to stochastic characteristics of transient fault occurrences and uncertainties in energy availability, enabling reliable and near-perpetual system operation in harsh environments has become a major design concern in energy harvesting real-time systems.

In this project, we first propose a temperature-aware task allocation scheme for multiprocessor systems, which exploits heterogeneous characteristics of real-time tasks to optimize system energy consumption and minimize processor peak temperature at the fine-grained level. We then design a task scheduling scheme for uniprocessor systems that jointly considers energy, temperature and reliability.

The energy consumed by individual processors is minimized without compromising the system reliability and thermal constraint.
In the runtime, a state-aware dynamic task scheduling scheme is developed to adapt system behavior to the harvesting power and system utilization.

Instead of decoupling energy and time, the proposed runtime scheme makes energy/time trade-off explorations for individual tasks and utilizes the exploration results to achieve near-perpetual operation of energy harvesting real-time systems.

 

Team composition

 

ECNU

  • Prof. Mingsong Chen
  • Prof. Tongquan Wei
  • Doctoral student: Yuan Yao

 

ENS of Lyon

  • Prof. Anne Benoit
  • Prof. Loris Marchal
  • Doctoral student: Changjiang Gou
Subject(s)