高级检索
    廖 彬, 于 炯, 孙 华, 年 梅. 基于存储结构重配置的分布式存储系统节能算法[J]. 计算机研究与发展, 2013, 50(1): 3-18.
    引用本文: 廖 彬, 于 炯, 孙 华, 年 梅. 基于存储结构重配置的分布式存储系统节能算法[J]. 计算机研究与发展, 2013, 50(1): 3-18.
    Liao Bin, Yu Jiong, Sun Hua, Nian Mei. Energy-Efficient Algorithms for Distributed Storage System Based on Data Storage Structure Reconfiguration[J]. Journal of Computer Research and Development, 2013, 50(1): 3-18.
    Citation: Liao Bin, Yu Jiong, Sun Hua, Nian Mei. Energy-Efficient Algorithms for Distributed Storage System Based on Data Storage Structure Reconfiguration[J]. Journal of Computer Research and Development, 2013, 50(1): 3-18.

    基于存储结构重配置的分布式存储系统节能算法

    Energy-Efficient Algorithms for Distributed Storage System Based on Data Storage Structure Reconfiguration

    • 摘要: 作为云计算底层核心基础设施,分布式数据存储系统是各种云计算服务的基础,是云计算重要的组成部分.然而随着系统规模的不断扩大以及设计时对能耗因素的忽略使其暴露出高能耗问题.由于存在数据可用性要求,使其并不能简单采用已有节能技术解决能耗问题,保证数据可用性成为设计分布式存储系统节能算法的前提.对系统与数据的可用性进行了建模,并通过对存储结构与机制、节点状态与数据块可用性之间的关系的研究提出构造数据块可用性度量矩阵解决数据可用性完全覆盖问题.定义了分布式存储系统的节能模型,将RACK划分为Active-Zone与Sleep-Zone两个存储区域,根据不同数据的访问频率与规律计算活动因子以配置数据的存储区域,通过数据中心负载规律适时对Sleep-Zone区域中的服务器进行休眠处理以达到节能的目的.实验结果表明:算法在适应数据访问规律与可用性保证前提下,能够提高分布式存储系统的能耗利用率,并且当系统负载与数据平均活动因子越低时算法节能效率越高.

       

      Abstract: As an underlying core infrastructure and important component of cloud computing, distributed storage system is the foundation of all kinds of cloud services or applications. However, with the expanding of system scale and energy consumption factors being ignored by its designers, the problem of high energy consumption is exposed. Because of data availability requirements, we cannot simply use the existing energy-saving technologies to solve the distributed storage system’s high energy consumption problem. To ensure all data’s availability is the premise of designing energy-efficient algorithms for distributed storage system. In this paper, we create a system and data availability model. After studying the storage structure and mechanism, the relationship between server’s status and data’s availability, the method to solve the problem of full covering by constructing a data’s availability measurement matrix is proposed. The energy saving model for distributed storage system is defined. The RACK is divided into Active-Zone and Sleep-Zone distinct storage area. According to the different data access frequency and regularity, we calculate the different activity factors for different data which decide how many replicas are stored in Active-Zone. The servers in Sleep-Zone are turn to sleep for energy saving while the work load of data center is low. Experimental results show that, the energy-efficient algorithms in this paper adapt the access rule and the availability of all the data in system, improve the energy-efficient of distributed storage system, and the algorithm is more efficient when the system’s work load and average activity factor is low.

       

    /

    返回文章
    返回