Open Access

ARTICLE

An OP-TEE Energy-Efficient Task Scheduling Approach Based on Mobile Application Characteristics

by Hai Wang^*, Xuan Hao, Shuo Ji^*, Jie Zheng, Yuhui Ma, Jianfeng Yang

State-Province Joint Engineering and Research Center of Advanced Networking and Intelligent Information Services, School of Information Science and Technology, Northwest University, Xi’an, 710127, China

* Corresponding Authors: Hai Wang. Email: ; Shuo Ji. Email:

(This article belongs to the Special Issue: Emerging Trends in Big Data Driven Edge Intelligence and In-Network Computing)

Intelligent Automation & Soft Computing 2023, 37(2), 1621-1635. https://doi.org/10.32604/iasc.2023.037898

Received 20 November 2022; Accepted 12 April 2023; Issue published 21 June 2023

Abstract

Trusted Execution Environment (TEE) is an important part of the security architecture of modern mobile devices, but its secure interaction process brings extra computing burden to mobile devices. This paper takes open portable trusted execution environment (OP-TEE) as the research object and deploys it to Raspberry Pi 3B, designs and implements a benchmark for OP-TEE, and analyzes its program characteristics. Furthermore, the application execution time, energy consumption and energy-delay product (EDP) are taken as the optimization objectives, and the central processing unit (CPU) frequency scheduling strategy of mobile devices is dynamically adjusted according to the characteristics of different applications through the combined model. The experimental result shows that compared with the default strategy, the scheduling method proposed in this paper saves 21.18% on average with the Line Regression-Decision Tree scheduling model with the shortest delay as the optimization objective. The Decision Tree-Support Vector Regression (SVR) scheduling model, which takes the lowest energy consumption as the optimization goal, saves 22% energy on average. The Decision Tree-K-Nearest Neighbor (KNN) scheduling model with the lowest EDP as the optimization objective optimizes about 33.9% on average.

---

Keywords

---