检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
机构地区:[1]中国科学技术大学计算机学院安徽省通信与软件重点实验室,合肥230027
出 处:《小型微型计算机系统》2013年第5期1021-1027,共7页Journal of Chinese Computer Systems
基 金:国家自然科学基金重点项目(60833005)资助
摘 要:闪存的擦后写特性,使其对小粒度随机更新为主的数据库应用,存在较大的更新时延.基于块内日志的存储管理模型提出了一种使用日志的方法,有效地解决了该问题.但是由于没有考虑数据访问的冷热特性,使得热擦除块合并操作非常频繁,同时它们采用的强制日志刷新策略导致闪存日志区存在严重的碎片问题.针对上述问题,本文提出一种基于数据冷热检测的双链表缓冲区算法DLPA,它根据数据的访问特性动态地分配日志页大小,可以有效减少擦除块合并操作,同时在日志刷新至闪存时,结合两种日志打包策略,有效地改善了日志区碎片问题.实验显示,该算法在增加少量存储开销的前提下,显著地优于现有算法.The erase-before-write characteristic of flash memory results in unacceptable update latency for small-sized random updates dominant database applications, like on-line transaction process. Some researchers have proposed an in-page logging approach to solve this problem. It flushes the update logs instead of the data pages themselves to flash when a physical write needs to be performed. But they do not consider the data access pattern. So some hot blocks will trigger a merge operation very frequently. At the same time, the mandatory log flushing strategy leads to serious fragmentation on the flash log region. In this paper, we present an effective buffer management algorithm called DLPA. It detects the data access pattern and dynamically allocates the log pages in the buffer. When the log pages are to be flushed, we perform a log padding operation to improve the log utilization of the log region. To verify our algorithm, we simulate a flash storage platform and run a series of TPC-C benchmark tests. The experiment results show our approach outperforms other solutions almost in all cases especially when the trace is skewed.
分 类 号:TP311[自动化与计算机技术—计算机软件与理论]
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:216.73.216.28