NVM Memory非揮發性記憶體(例如 NAND 快閃記憶體)中的安全資料刪除在資訊安全的考慮下，對於防止潛在的敏感資料洩露是不可忽視的議題。選擇安全刪除方法除了須考慮所涉及資料的敏感程度，同時在遵循SSD特性下實踐資料安全刪除。本研究中，我們提出了一種基於 LevelDB和 OCSSD 的新方法，針對寫入密集型資料庫，採用WiscKey將鍵值分離儲存的機制並設計了一個名為「Version Table」的溝通中樞，也就是作為資料庫中key和相關各版本的value log所在的中介。此外，當使用者操作「Erase」命令時，會同時觸發Data Snitization。因此當啟動Garbage Collection時，會優先考慮包含更高比例Sanitized data的block作為刪除對象，在不嚴重影響效能的前提下進行對資料安全刪除的操作。

Secure data deletion within non-volatile memory, such as NAND flash memory is necessary to guard sensitive data and thwart potential data breaches. The selection of a secure deletion method should align with the sensitivity of the data in question, while adhering to the best practices considering the characteristic of SSD. In this study, we propose a novel approach based on LSM-tree, LevelDB and OCSSD for write-intensive databases. To begin with, we adopted the strategy of WiscKey, which stores key and value in LSM-tree and value log, respectively. Further, we use a dedicated communication hub known as the "Version Table." This Version Table acts as an intermediary between the database, where keys and associated value logs reside. Moreover, when a user initiates a "Delete" operation, data sanitation procedures are triggered concurrently. Consequently, when the Garbage Collection process is invoked, blocks housing a higher proportion of sanitized data are accorded higher priority in the space reclamation process, optimizing data security and resource utilization.

Contents
Abstract (Chinese) I
Acknowledgements (Chinese) II
Abstract III
Contents IV
List of Figures VI
List of Tables VII
List of Algorithms VIII
1 Introduction 1
2 Background 2
2.1 Key-Value Storage . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.1 Security Deletion of Key-Value Pair . . . . . . . . . . . . . . 2
2.2 LevelDB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.1 WiscKey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Open Channel SSDs (OCSSD) . . . . . . . . . . . . . . . . . . . . . 5
2.4 Hot-Cold separation . . . . . . . . . . . . . . . . . . . . . . . . . . 8
IV
3 Methodology 9
3.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Version Tracking Strategy . . . . . . . . . . . . . . . . . . . . . . . 9
3.3 Hot-Cold Separation . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Data Sanitization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4 Experimental Results 18
4.1 Yahoo! Cloud Serving Benchmark (YCSB) . . . . . . . . . . . . . . 19
4.2 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.3.1 Impact of Model Architecture . . . . . . . . . . . . . . . . . 20
4.3.2 Total Write Page Quantities . . . . . . . . . . . . . . . . . . 20
4.3.3 Garbage Collection and Number of Live Page Copying . . . 21
5 Conclusion 23
Bibliography 24

[1] Justin Sheehy and David Smith. Bitcask – a log-structured hash table for fast key/value data. Basho White Paper, 2010.
[2] J. Dean and S. Ghemawat. Leveidb. [online] Available:
http://code.google.com/p/leveldb, 2011.
[3] Andrea C. Arpaci-Dusseau Lanyue Lu, Thanumalayan Sankaranarayana Pillai and Madison. Remzi H. Arpaci-Dusseau. University of Wisconsin. Wisckey:Separating keys from values in ssd-conscious storage. fast16 2016, pages 1–8,2008.
[4] T.-W. Kuo J.-W. Hsieh and L.-P. Chang. Efficient identification of hot data for flash memory storage systems. ACM Transactions on Storage,vol. 2, (1),2006.
[5] Dongchul Park and David H.C. Du. Hot and cold data identification for flash memory using multiple bloom filters. FAST 11, page 1–11, 2011.
[6] Brian F. Cooper, Adam Silberstein, Erwin Tam, Raghu Ramakrishnan, and Russell Sears. Benchmarking cloud serving systems with ycsb. In Proc. of SoCC, 2010.