Towards Flash Disk Use in Databases - Keeping Performance While Saving Energy?
Towards Flash Disk Use in Databases - Keeping Performance While Saving Energy?
复制标题
DOI：
--
复制
发表时间：
2009
期刊：
影响因子：
--
通讯作者：
T. Härder;Karsten Schmidt;Y. Ou;Sebastian Bächle
中科院分区：
文献类型：
--
作者：
T. Härder;Karsten Schmidt;Y. Ou;Sebastian Bächle
处理信息时的绿色计算或节能主要被认为是处理器开发的任务。然而，我们主张采取整体方法将功耗降至最低。我们探讨的潜力NAND闪存相比，磁盘的DBMS为基础的架构。出于这个原因，我们试图确定在设备级别和文件系统接口的存储类型的IO性能，然后我们接近最重要的问题，DBMS应用程序如何利用闪存的性能特性。此外，基于我们的XTC系统中的存储结构和相关的基准测试运行的分析，我们讨论了如何IO密集型DBMS算法必须进行调整，以采取大部分的闪存技术，这是进入服务器领域的戏剧性的步伐。最终目标是在保持与基于磁盘的数据库服务器相当的性能的同时，大幅提高能源效率。
Green computing or energy saving when processing information is primarily considered a task of processor development. We, however, advocate that a holistic approach is necessary to reduce power consumption to a minimum. We explore the potential of NAND flash memory in comparison to magnetic disks for DBMS-based architectures. For this reason, we attempt to identify the IO performance of both storage types at the device level and file system interface, before we approach the most important question how DBMS applications can take advantage of the flash performance characteristics. Furthermore, based on an analysis of storage structures and related benchmark runs in our XTC system, we discuss how IO-intensive DBMS algorithms have to be adjusted to take most out of the flash technology which is entering the server area with dramatic pace. The ultimate goal is to substantially improve energy efficiency while comparable performance as in disk-based DB servers is maintained.