An Energy-Aware Dynamic Algorithm Based on Variable Interval DVFS Technology

Abstract

. Nowadays, embedded system has been integrated into the various fields of people's life. The contradiction between energy efficiency and time efficiency is more and more seriously. A system with DVFS module can relieve this contradiction. But a lot of researches witch concern about it has kinds of defects. This paper designed a new energy consumption model for a DVFS system and a new simple multi-phase task module for the task generating and processing. Then a dynamic DVFS method was designed to run in our platform with kinds of different types of tasks. We introduced a mechanism with dynamic frequency update interval into our method, and proved that in most cases it can achieve better energy efficiency than other two algorithms which use a fixed frequency update interval. Meanwhile, it can hold the similar time efficiency to the algorithm which makes the best time efficiency in them.

A Time Reservation Dynamic Algorithm Based on the Variable Frequency Interval DVFS Technology

“…Based on the research work [7] to [11], our previous work [12] improved the algorithm based on variable interval DVFS technology. We have improved the theoretical models in DVFS research and the corresponding algorithm.…”

“…So, we need adjust the frequency interval to the fast mode and track the frequency bound in the task to minimize the possibility of task timeout. In order to test the actual performance of TPD algorithm, we use the same experimental platform and set of tasks as the one in the DVI algorithm in [12], so it's easily to compare the ratio of energy consumption and task timeout with both algorithms. The task sets are as follows: By executing these tasks set in table 1 in our experimental platform, the result is as follows: The experiment shows that TRD algorithm can achieve a better response time than DVI algorithm in task processing.…”

“…Then, Reference [12] designed a DVI algorithm by switching the interval of frequency adjustment to change the system frequency adjusting speed. Fast frequency interval can enhance the adaptability of CPU on task frequency bound mutation, but it will generate higher energy costs.…”

“…Reference [12] uses a multi-phase task model to show frequency bound characteristics of the task. The researchers can determine the grain size of each task phase according to their own requirements, for a finer grain of task phase, it can reflect the more detail features of the task.…”