项目名称: 基于任务行为特征分析的热敏感操作系统技术研究
项目编号: No.61272131
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 李曦
作者单位: 中国科学技术大学
项目金额: 68万元
中文摘要: 计算机系统热敏感控制问题是近年来学术界研究的一个热点问题。本课题以平衡系统的性能,公平性,功耗,温度等关键参数为目标,结合操作系统和体系结构理论,研究任务行为特征分析、线程分组管理和保持功能部件状态亲等关键理论和技术,在操作系统中实现热敏感的任务和资源管理框架。与已有研究的差异在于:1.以任务的行为特征分析为研究重点,首次提出"功能事件"概念刻画线程的冷热特征和线程间的依赖关系,在此基础上抽象出任务模式;2.提出新的"线程组"的概念,按不同类型的任务模式进行线程分组,分配CPU和内存等硬件资源;3.首次提出"状态亲和性"的概念,在保证系统性能和线程公平性的前提下,降低系统关键部件的功耗和温度;4.基于线程分组管理和状态亲和性思想,设计热敏感的操作系统任务调度和资源分配算法。本项目对系统级热敏感控制理论和技术研究有重要推动和贡献,对提升我国在该领域的创新能力有重要的意义。
中文关键词: 任务行为特征;操作系统;热敏感;组调度;
英文摘要: Thermal-aware control problem is a hot academic research issue in computer system in recent years. The topic aims to balance the system performance, fairness, power, temperature and other key parameters, combining with the operating system and architecture theory, researching the characteristics of the task behavior, thread grouping management and maintaining the state of affinity of the features and other critical theoretical and technology, achieving thermal-aware task and resource management in the operating system. The difference with the past researches is that: 1. Focused on the characteristics of the task behavior, first proposed the concept of the "feature event", portrayed the hot and cold characteristics of the thread and dependencies between the threads, abstracted task pattern according to feature event; 2. Proposed a new "thread group" concept, partitioned threads to thread group according to the characteristics of the task behavior, allocating hardware resources such as CPU and memory based on thread group; 3. First proposed the concept of "state affinity" of the features, reducing system power and temperature under the premise of ensuring system performance and thread fairness; 4. Based on the ideology of thread grouping management and maintaining state affinity, design task scheduling and resourc
英文关键词: characteristics of the task behavior;operation system;thermal-aware;group scheduling;