臺灣博碩士論文加值系統

近幾年來快閃記憶體(Flash memory)已被廣泛的應用在各種運算環境中，藉由快閃記憶體所組成的固態硬碟(SSD)也逐漸取代傳統硬碟(HDD)，成為運算系統中的主要儲存裝置。雖然使用更加微縮的製程可以提升單位面積的容量以及增加存取效能，但也同時減低各個Cell的耐用度，進而造成了整體SSD的壽命下降。過去已經提出許多技術用來提升SSD的壽命與效能。重複資料刪除(Data Deduplication)透過刪除重複資料減少資料寫入來提升寫入速度，進而延長SSD的壽命；資料複製(Data Replication)透過複製重要資料來提升資料可靠度。但過度的重複資料刪除可能會將重要的資料備份刪除，造成重要資料的可靠度下降；過度的資料複製可能會複製過多不重要的資料，進而傷害SSD的壽命。本研究的關鍵議題在於針對重複資料刪除與資料複製在SSD中的平衡。當資料寫入時，透過in-line的重複資料刪除，在資料寫入SSD前刪除重複資料，並透過同步偵測並複製重要資料，進而達到重要資料的可靠度提升、SSD的壽命延長與資料存取速度的提升。

Flash memory has been widely used as a storage medium for portable device, personal computer and data center due to non-volatility, high performance, low power consumption and shock resistance. Some techniques have been proposed to achieve better lifetime or performance of flash memory. Data deduplication has been proposed for reducing the duplicated data to reduce the write traffic to extend the lifetime of flash memory and achieving higher performance. Data Replication replicated the data to achieve higher reliability. However, blindly doing data deduplication might deduplicated the critical data that will damage the reliability of the storage. On the other hand, blindly doing data replication might replicated too many data that leads flash memory wear out earlier than usually. In this paper, we proposed a novel flash translation layer algorithm Dr. FTL by using data deduplication and data replication, it focus on lifetime improvement of critical data and enhanced performance as main objectives. Dr. FTL detect the duplication data and doing data duplication when the request arrival, thus reduce write traffic to SSD. Meanwhile, we detect the critical data and replicated these data to achieve higher reliability of the critical data. The experiments shows that Dr. FTL convincingly for the general workloads. And it improves the lifetime of critical data by up to 32% in general workloads. For performance concern, Dr. FTL also achieves better performance than Normal FTL. These results shows that Dr. FTL is better to use for the applications which need to achieving higher reliability of critical data and higher performance.

1 Introduction
2 Background and Motivation
2.1 Flash Memory
2.2 Data Deduplication
2.3 Data Replication
2.4 Critical Data
2.5 Motivation
3 The Design of Dr. FTL
3.1 Overview of Dr. FTL
3.2 Mapping Table
3.3 Data Replication
3.4 Garbage Collection
3.5 Channel Selection
4 Performance Evaluation
4.1 Experimental Setup
4.2 Experimental Results
4.2.1 Lifetime Improvement of Critical Data
4.2.2 Replication Threshold Analysis
4.3 Performance Improvement
4.3.1 Total Write Page Count
4.3.2 Total Channel Selection Count
5 Conclusion

[1] M-System, \Two technologies compared: Nor vs nand," July 2003, White Paper. [Online]. Available: http://maltiel-consulting.com/
[2] T. Hara, \A 146-mm2 8-gb multi-level nand flash memory with 70-nm cmos technology," IEEE Solid-State Circuits Society, vol. 41, pp. 161{169, 2006.
[3] Y. Li, \A 16gb 3b/ cell nand flash memory in 56nm with 8mb/s write rate," in IEEE Solid-State Circuits Conference, 2008.
[4] L.-P. Chang, \A hybrid approach to nand-flash-based solid-state disks," IEEE Trans. Computers, vol. 59, no. 10, pp. 1337{1349, 2010.
[5] F. Chen, T. Luo, and X. Zhang, \Caftl: A content-aware flash translation layer enhancing the lifespan of flash memory based solid state drives." USENIX, 2011, pp. 77{90.
[6] S. Lee, T. Kim, J. Park, and J. Kim, \An integrated approach for managing the lifetime of flash-based ssds," in DATE, 2013, pp. 1522{1525.
[7] R. Manjunath and T. Xie, \Dynamic data replication on flash ssd assisted video-on-demand servers." IEEE Computing, Networking and Communications, 2012, pp. 502{506.
[8] N. Altiparmak and A. S. Tosun, \Replication based QoS framework for flash arrays," in CLUSTER, 2012, pp. 182{190.42
[9] Y.-H. Chang and T.-W. Kuo, \A reliable mtd design for mlc flash-memory storage systems," in EMSOFT, 2010, pp. 179{188.
[10] Y.-B. Chang and L.-P. Chang, \A self-balancing striping scheme for nand flash storage systems," in SAC, 2008, pp. 1715{1719.
[11] P.-H. Yeh, \A multi-channel management scheme optimized for read through-put," Master's thesis, the Department of Computer Science and Information
Engineering, National Taiwan University of Science and Technology, Taipei,Taiwan, 2013.
[12] A. Gupta, Y. Kim, and B. Urgaonkar, \Dftl: a flash translation layer employing demand-based selective caching of page-level address mappings," in ASPLOS, 2009, pp. 229{240.
[13] A. Kawaguchi, S. Nishioka, and H. Motoda, \A flash-memory based file system," in USENIX Winter, 1995, pp. 155{164.
[14] Y. Hu, H. Jiang, D. Feng, L. Tian, H. Luo, and C. Ren, \Exploring and exploiting the multilevel parallelism inside ssds for improved performance and endurance," IEEE Trans. Computers, vol. 62, no. 6, pp. 1141{1155, 2013.
[15] \Samsung electronics company k9gag08u0m 2g * 8 bit mlc nand flash memory data sheet (preliminary)," 2006.
[16] L.-P. Chang and C.-D. Du, \Design and implementation of an e fficient wear-leveling algorithm for solid-state-disk micro controllers," ACM Trans. Design Autom. Electr. Syst., vol. 15, no. 1, 2009.
[17] U. T. Repository, \Oltp applications of two financial institutions." [Online]. Available: http://traces.cs.umass.edu/