中文摘要：

动态电压频率缩放（DVFS）技术是当前最有效的功耗调节手段之一．本文首先分析现有DVFS技术存在的不足，指出限制DVFS技术高效运用的核心因素；基于现有低效的方式我们提出一种基于任务行为分析的DVFS机制（TC-DVFS）．其具有三个层次：一、采集任务的系统调用信息；二、识别任务的关键系统调用，并以关键系统调用刻画任务行为；三、根据任务行为构建特征库，并以任务的特征库来指导DVFS．我们将TC-DVFS添加到linux内核中，并在intel-core2处理器平台上对不同类型的应用任务进行性能与功耗测试．结果显示TC-DVFS总体获得10％的性能提升，并降低5％调频失效率和5％的系统能耗．

英文摘要：

The dynamic voltage and frequency scaling （DVFS） is one of the most effective mechanism for power management. Firstly, we point out the most shortcomings of the current DVFS mechanisms after detailly analyzing; so we propose a task behavior based DVFS mechanism （TC-DVFS）. TC-DVFS has three levels： First, gather the system call information of the task. Second, distinguish the critical system calls and characterize task behavior with them; Third, construct a feature database based on task behavior and guide DVFS according the database. Adding TC-DVFS into linux kernel, experimental results show TC-DVFS reduce 5% miss rate of the correct frequency scaling and 5% power of the system, and also improve 10% performance of the system.