臺灣博碩士論文加值系統

SSD（固態硬碟）是已被廣泛使用的非揮發性記憶體，因為與HDD（硬碟機）相比，其能耗低，重量輕，尺寸更小。但是，由於它具有批量清除和一次寫入等特性，高效的FTL（快閃記憶體轉換層）非常重要。隨著這個資訊科技時代的技術不斷發展，大量且頻繁的數據輸入輸出越來越重要。為了減少大量存儲的操作所花的時間以及記憶體的使用壽命，可以利用重複數據刪除等技術，以減少記憶體中的讀寫次數。一些研究，如CAFTL（內容感知快閃記憶體轉換層）已經在FTL中應用了重複數據刪除的技術，以減少快閃記憶體中的清除次數。我們提出一種在快閃記憶體中進行重複數據刪除的方法，優化GC（垃圾回收）處理指紋儲存以達成減少GC overhead。實驗在SSD模擬器中進行，
對照CAFTL與我們的方法，發現讀與寫的操作時間都有所改善。
關鍵詞： 快閃記憶體、重複數據刪除、指紋存儲

SSD (Solid State Drives) has been widely used as a non-volatile memory since compare with hard disk drive (HDD) its low energy consumption, low weight, and smaller size. However, due to it’s characteristic like bulk-erasing and write-once, an efficient Flash Translation Layer (FTL) is important. As the technology of the information age has developed further and further, large and frequent data saving loading is more and more crucial. To handle the large storage operation, technics such as deduplication had been introduced, reducing the read/write count in memory. Some research such as Content-Aware Flash Translation Layer (CAFTL) has handled the issue to apply deduplication handling in FTL, to reduce erase cycle in flash memory. However, handling deduplication such as maintaining fingerprint store cause extra operation time especially during GC (Garbage Collection). This paper will introduce a solution to deduplication in flash memory by optimizing GC fingerprint-stores, in order to reduce overhead during GC. Experiment is conducted in an SSD simulator, simulate CAFTL and our work, and found improvements in both read and write operation time.
Keyword: Flash Memory, Deduplication, Fingerprint Store

摘 要 iii
Abstract iv
Lists of Tables vii
1. Introduction 1
2. Background and Motivation 3
2.1 Characteristic of Flash Memory 3
2.1.1 Life-span (Erase cycle) 3
2.1.2 Garbage Collection 4
2.2 Deduplication 5
2.2.1 Hashing 5
2.2.2 Fingerprint Store 5
2.3 CAFTL 6
2.3.1 Two-level indirect mapping 6
2.4 Motivation 8
3. Method 9
3.1 Fingerprint store and Invalid Cache 10
3.1.1 Fingerprint store 10
3.1.2 Invalid Cache 11
3.2 Duplication 11
3.3 Garbage Collection (GC) 12
4. Performance Evaluation 14
4.1 Experiment Setup 14
4.1.1 SSD Simulator and Configuration 14
4.1.2 Trace Data 14
4.2 Experiment Result 15
5. Conclusion 18
6. Reference 19

[1]Chang, Y. H., Hsieh, J. W., & Kuo, T. W. (2007, June). Endurance enhancement of flash-memory storage systems: An efficient static wear leveling design. In Proceedings of the 44th annual Design Automation Conference (pp. 212-217).
[2]Chen, F., Luo, T., & Zhang, X. (2011, February). CAFTL: A Content-Aware Flash Translation Layer Enhancing the Lifespan of Flash Memory based Solid State Drives. In FAST (Vol. 11, pp. 77-90).
[3]Eastlake 3rd, D., & Jones, P. (2001). US secure hash algorithm 1 (SHA1) (No. rfc3174).
[4]Goodson, G., & Iyer, R. (2010, January). Design tradeoffs in a flash translation layer. In Proceedings of Workshop on the Use of Emerging Storage and Memory Technologies.
[5]Hu, X. Y., Eleftheriou, E., Haas, R., Iliadis, I., & Pletka, R. (2009, May). Write amplification analysis in flash-based solid state drives. In Proceedings of SYSTOR 2009: The Israeli Experimental Systems Conference (pp. 1-9).
[6]Wang, S., He, C., Tang, J., Lu, X., Shen, C., Yu, H., ... & Zhang, G. (2019). New floating gate memory with excellent retention characteristics. Advanced Electronic Materials, 5(4), 1800726.
[7]Yadgar, G., Gabel, M., Jaffer, S., & Schroeder, B. (2021). SSD-based workload characteristics and their performance implications. ACM Transactions on Storage (TOS), 17(1), 1-26.
[8]Zhou, D., Pan, W., Wang, W., & Xie, T. (2015, October). I/O characteristics of smartphone applications and their implications for eMMC design. In 2015 IEEE International Symposium on Workload Characterization (pp. 12-21). IEEE.