Precision versusPerformance: Optimizing Swapping Mechanisms in Multi-Mode NVM

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：133

、訪客IP：3.16.207.215

姓名

吳苡菡(Yi-Han Wu) 查詢紙本館藏

畢業系所

資訊工程學系

論文名稱

(Precision versusPerformance: Optimizing Swapping Mechanisms in Multi-Mode NVM)

相關論文

★ 重新思考虛擬記憶體管理的方式以開放通道式固態硬碟最大限度地減少深度學習推薦系統演算法的讀寫流量	★ 開啟製程相似檢查方法在組裝超級塊上以最小化額外的寫入延遲
★ LaDy: Enabling Locality-aware Deduplication Technology on Shingled Magnetic Recording Drives	★ On Minimizing Writing Overhead to Establish a Low-latency LSM-tree on Skyrmion-based Racetrack Memory
★ WABE: Rethinking B-epsilon-tree to Minimize Write-amplification on NAND Flash Memory	★ Rethinking Bϵ tree Indexing Structure over NVM with the Support of Multi-write Modes
★ Prophet’s Insight: Unleashing Deduplication System Performance in Multi-tier Storage Systems	★ Freeing the Power of High Parallelism: Accelerating the Bϵ-Tree Indexing Scheme Performance on Open-Channel SSD
★ Applying Content-Defined Chunking to OCSSD-based Deduplication Systems	★ GraLoc: Preserving Graph Locality to Minimize Read and Write Amplification on NAND Flash Memory
★ On Minimizing Writing Overhead to Establish a Low-latency LSM-tree on Skyrmion-based Racetrack Memory	★ Planting a Forest in Sky: Harnessing Parallelism in Skyrmion Racetrack Memory for Efficient Random Forest Data Placement
★ LaGRange: Locality-aware Graph with Range Maintenance

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

至系統瀏覽論文 (2027-7-20以後開放)

摘要(中)

非揮發性記憶體（Non-volatile memory）有望取代動態隨機存取記憶體（Dynamic Random Access Memory）和 NAND 快閃記憶體，成為下一代的主要記憶體和二級存儲裝置。儘管目前因為 NVM 的壽命限制，使它無法成為主流的主記憶體，但非揮發性記憶體的大存儲容量和高單元密度，使其成為二級存儲裝置，或是作為 NAND 快閃記憶體之上硬體快取的理想選擇。
隨著非揮發性記憶體在二級主要記憶體和虛擬記憶體系統中的使用增加，重新評估當前的虛擬記憶體管理策略至關重要，因為現有的策略通常未能最大化非揮發性記憶體的優勢。多層單元（Multi-level-cell）非揮發性記憶體支援多種寫入模式，具有不同的成本和數據保留時間。有效的虛擬記憶體管理應該考慮到程序的類型，以選擇最佳的寫入模式，從而最小化非揮發性記憶體的能耗。具體而言，它應根據精確 (precise) 和近似 (approxi mate) 程序的特性選擇相應的寫入模式。
本研究介紹了 eSwap，一種性能高效的交換機制，選擇適當的寫入模式和錯誤校正碼（Error-correction code）成本，以管理多層單元非揮發性記憶體上的精確和近似程序。實驗結果表明，與最先進的解決方案相比，eS wap 能夠提高 20% 的能耗和 40% 的系統性能。

摘要(英)

Non-volatile memory (NVM) is poised to replace Dynamic Random Access Memory (DRAM) and NAND flash memory as the next-generation main memory and secondary storage. Despite current lifetime limitations that hinder its immediate use as main memory, NVM’s large storage capacity and high cell density make it ideal for secondary memory or as a hardware cache on top of NAND flash memory. As NVM usage grows in secondary main memory and virtual memory systems, reevaluating current virtual memory management strategies is crucial, as they often fail to maximize NVM’s benefits. Multi-level-cell (MLC) NVM supports various writing modes with different costs and data retention times. Effective virtual memory management should consider the type of process to select the optimal writing mode, thus minimizing NVM energy consumption. Specifically, it should differentiate between precise and approximate processes when writing to MLC NVM. This study introduces eSwap, a performance-efficient swapping mechanism that chooses the appropriate writing mode and error-correction code (ECC) cost for managing precise and approximate processes on MLC NVM. Experimental results indicate that eSwap enhances energy consumption and system performance by 20\% and 40\%, respectively, compared to state-of-the-art solutions.

關鍵字(中)

★ 非揮發性記憶體
★ 交換空間
★ 近似記憶體
★ 多寫入模式

關鍵字(英)

★ non-volatile memory
★ swap space
★ approximate memory
★ multi-write modes

論文目次

Contents
Abstract ii
1 Introduction 1
2 Technical Background & Motivation 4
2.1 Approximate Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3 Multi-write Swapping Strategy 8
3.1 eSwap Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Scrubbing Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
short zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
long zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
heavy zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 Allocating page to Scrubbing zone . . . . . . . . . . . . . . . . . . . . 11
Write hot page . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Approximate process and read cold page . . . . . . . . . . . . 11
Approximate process and read hot page . . . . . . . . . . . . . 11
precise process and read hot page . . . . . . . . . . . . . . . . . 12
precise process and read cold page . . . . . . . . . . . . . . . . 12
3.4 Page state determination . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.5 Aging Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.6 Rules of selecting victim page . . . . . . . . . . . . . . . . . . . . . . . 14
3.6.1 Victim page priority . . . . . . . . . . . . . . . . . . . . . . . . . 15
iii
3.6.2 Scrubbing desired times table . . . . . . . . . . . . . . . . . . . 15
3.7 Overhead Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4 Experiments 17
4.1 Environmental Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.1.1 Error Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Evaluation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Comparison of Energy Consumption . . . . . . . . . . . . . . . 21
Comparison of Latency . . . . . . . . . . . . . . . . . . . . . . . 21
Comparison of Space Overhead . . . . . . . . . . . . . . . . . . 22
Comparison of ECC Activation times . . . . . . . . . . . . . . . 23
5 Concluding Remarks 26
Bibliography 27

參考文獻

Bibliography
[1] Mingzhe Zhang et al. “Balancing performance and lifetime of MLC PCM by
using a region retention monitor”. In: 2017 IEEE International Symposium on High
Performance Computer Architecture (HPCA). IEEE. 2017, pp. 385–396.
[2] Zitao Chen et al. “Binfi: An efficient fault injector for safety-critical machine
learning systems”. In: Proceedings of the International Conference for High Perfor-
mance Computing, Networking, Storage and Analysis. 2019, pp. 1–23.
[3] Giulia Stazi et al. “Impact of approximate memory data allocation on a h. 264
software video encoder”. In: International Conference on High Performance Com-
puting. Springer. 2018, pp. 545–553.
[4] Chen Pan et al. “3M-PCM: Exploiting multiple write modes MLC phase change
main memory in embedded systems”. In: (2014), pp. 1–10. DOI: 10 . 1145 /
2656075.2656076.
[5] Sudarsun Kannan et al. “Optimizing Checkpoints Using NVM as Virtual Mem-
ory”. In: 2013 IEEE 27th International Symposium on Parallel and Distributed Pro-
cessing. 2013, pp. 29–40. DOI: 10.1109/IPDPS.2013.69.
[6] Amro Awad, Sergey Blagodurov, and Yan Solihin. “Write-aware management
of nvm-based memory extensions”. In: Proceedings of the 2016 International Con-
ference on Supercomputing. 2016, pp. 1–12.
[7] Yunjoo Park and Hyokyung Bahn. “Management of virtual memory systems
under high performance PCM-based swap devices”. In: 2015 IEEE 39th Annual
Computer Software and Applications Conference. Vol. 2. IEEE. 2015, pp. 764–772.
[8] Qingan Li et al. “Compiler directed write-mode selection for high performance
low power volatile PCM”. In: (2013), pp. 101–110.
[9] Keni Qiu, Qingan Li, and Chun Jason Xue. “Write mode aware loop tiling for
high performance low power volatile PCM”. In: Proceedings of the 51st Annual
Design Automation Conference. 2014, pp. 1–6.
[10] Remzi H. Arpaci-Dusseau and Andrea C. Arpaci-Dusseau. Operating Systems:
Three Easy Pieces. 1.10. Arpaci-Dusseau Books, 2023.
[11] Wei Xu, Jibang Liu, and Tong Zhang. “Data manipulation techniques to reduce
phase change memory write energy”. In: (2009), pp. 237–242.
[12] Yue Li, Anxiao Andrew Jiang, and Jehoshua Bruck. “Multi-Phase Scrubbing for
Tolerating Resistance Drift in Phase-Change Memories”. In: ().

指導教授

陳增益(Tseng-Yi Chen)

審核日期

2024-7-23

推文