An SSD-based disk cache architecture and its application to data system of data center of Meridian Project
-
摘要: 固态盘作为内存和磁盘之间的二级磁盘缓存运用于服务器存储层次结构中, 以满足空间科学研究领域日益增长的应用级I/O请求. 然而, 当前主流混合 存储架构未能充分发挥固态盘优势, 也没有考虑其寿命损耗. 因此, 本文提 出一种联合固态盘、磁盘的RAF (Random-Access First)混合存储架构, 在 提高系统性能的同时可延长固态盘寿命. 其主要思路是通过序列探测技术, 区 分负载中的随机访问和连续访问, 并将随机和顺序访问请求重定向到固态盘 和磁盘系统分别处理. RAF的原型已在Linux内核2.6.30.10中实现. 基于子 午工程真实业务数据及模拟数据的实验结果表明, RAF与当前的主流架 构Flash Cache相比, 在多种负载情况下可提升系统响应时间17%, 同时减少 固态盘老化率53%, 提高了存储系统的整体性能.Abstract: Solid-State Drivers (SSD) are integrated in server storage hierarchy as a second tier of disk cache between DRAM and disks for caching more data from disks to meet the increasingly intensive I/O demands. Unfortunately, available state-of-art hybrid storage architectures cannot fully exploit SSDs' potentials and shorten their lifetime. In this paper, RAF (Random Access First), a hybrid storage architecture that combines an SSD based disk cache and a disk drive subsystem, is proposed. RAF focuses on improving the system performance while extending the lifetime of SSD through providing priority to caching random-access data. Random-access and sequential-access data are identified by sequence detection scheme, and they will be served by SSD and Hard Disk Drive (HDD) respectively. RAF is implemented in Linux kernel 2.6.30.10. The results of experiments to apply it to the simulation and real data systems of Data Center of Meridian Project show that RAF can improve the performance by an average rate of 17% and reduce the flash wear by 53% compared with the state-of-art Flash Cache architecture.
-
Key words:
- SSD /
- Disk cache /
- Sequence detection
-
[1] Kgil T, Roberts D, Mudge T. Improving NAND Flash Based Disk Caches[C]//Proceedings of the 35th International Symposium on Computer Architecture. Beijing: IEEE Computer Society Washington, 2008. 327-338 [2] Kgil T, Mudge T. FlashCache: A NAND flash memory file cache for low power web servers[C]//CASES'06: Proceedings of the 2006 International Conference on Compilers, Architecture and Synthesis for Embedded Systems. Seoul: ACM New York, 2006. 103-112 [3] Wu M, Zwaenepoel W. eNVy: A non-volatile, main memory storage system[C]//Proceedings of the 6th International Conference on Architectural Support for Programming Languages and Operating Systems. San Jose: ACM New York, 1994. 86-97 [4] Narayanan D, Thereska E, Donnelly A, et al. Migrating server storage to SSDs: Analysis of tradeoffs[C]//Proceedings of the 4th ACM European conference on Computer systems. Nuremberg, Germany: ACM New York, 2009. 145-158 [5] Lee S, Moon B. Design of flash-based DBMS: An in-page logging approach[C]//Proceedings of the 2007 ACM SIGMOD International Conference on Management of Data. Beijing: ACM New York, 2007. 55-66 [6] Gupta A, Kim Y, Urgaonkar B. DFTL: A flash translation layer employing demand-based selective caching of page-level address mappings[C]//Proceeding of the 14th International Conference on Architectural Support for Programming Languages and Operating Systems. Washington DC: ACM New York, 2009. 229-240 [7] Uysal M, Merchant A, Alvarez G A. Using MEMS-based storage in disk arrays[C]//Proceedings of the 2nd USENIX Conference on File and Storage Technologies. San Francisco, CA: The USENIX Association, 2003. 89-101 [8] Marsh B, Douglis F, Krishnan P. Flash memory file caching for mobile computers[C]//Proceedings of the 27 Hawaii International Conference on System Sciences. Maui, HI: IEEE Computer Society, 1994. 451-460 -
点击查看大图
计量
- 文章访问数: 1718
- HTML全文浏览量: 10
- PDF下载量: 894
- 被引次数: 0
下载:
