Journal of Internet of Things and Convergence (사물인터넷융복합논문지)

The Korea Internet of Things Society (한국사물인터넷학회)
권호 표지
DOI QR코드

DOI QR Code

A Design of Hot Data Classifier and Management Scheme for An Efficient Resource Management

효율적인 자원 관리를 위한 핫데이터 분류기 및 관리 정책 설계

  • Hyun-Seob Lee (Division of Computer Engineering, Baekseok University)
  • 이현섭 (백석대학교 컴퓨터공학부)
  • Received : 2023.08.08
  • Accepted : 2023.10.09
  • Published : 2023.10.31
Download PDF
⟨ Previous Next ⟩

Abstract

Hot data is frequently used data that affects the performance and lifespan of a storage system. Therefore, classifying and managing the identified hot data is important for the performance and lifetime of storage devices, especially for NAND flash memory, which is characterized by rapid degradation after repeated intensive write operations to specific blocks. In this paper, we design a hot data classifier and management scheme for an efficient resource management in flash memory-based storage devices by benchmarking traces extracted from enterprise servers. Also, we propose a method to measure the frequency of hot data using multiple hashes, and to improve the performance and lifetime of the storage system by classifying the identified hot data and managing it on degradation-resistant media. Finally, we demonstrate the effectiveness of the proposed method by measuring the increased performance and lifetime through experiments.

핫데이터는 저장시스템의 성능과 수명에 영향을 주는 사용 빈도가 높은 데이터이다. 따라서 식별된 핫데이터를 분류하여 관리하는 것은 저장장치의 성능과 수명을 위해 중요한 일이다. 특히 특정 블록에 집중적인 쓰기 동작을 반복하면 열화 현상이 급격하게 진행되는 특징을 가지고 있는 낸드 플래시 메모리는 이러한 핫데이터를 구분하여 관리하는 것이 성능과 수명을 향상하기 위해 필요하다. 본 논문에서는 엔터프라이즈 서버에서 추출한 트레이스를 벤치마킹하여 플래시 메모리 기반 저장장치에서 자원을 효율적으로 관리하는데 필요한 핫데이터 분류기 및 관리 정책을 설계한다. 또한. 멀티 해시를 이용하여 핫데이터의 빈도를 측정하고, 식별된 핫데이터를 구분하여 열화에 강한 미디어에 관리하는 기법을 통해 저장시스템의 성능과 수명을 향상시키는 방법을 제안한다. 마지막으로 실험을 통해 증가한 성능과 수명을 측정하여 제안하는 방법의 효과를 증명한다.

Keywords

Acknowledgement

This paper was supported by 2023 Baekseok University Research Fund

References

  1. Q.Luo, X.Fang, Y.Sun, J.Ai and C.Yang, "Self-learning hot data prediction: Where echo state network meets NAND flash memories," IEEE Transactions on Circuits and Systems I: Regular Papers, Vol.67, No.3, pp.939-950, 2020. https://doi.org/10.1109/TCSI.2019.2960015
  2. F.Yu, and H.Yan, "An Efficient Hot-Cold Data Separation Garbage Collection Algorithm Based on Logical Interval in NAND Flash-based Consumer Electronics," IEEE Transactions on Consumer Electronics, Vol.69, No.3, pp.431-440, 2023. https://doi.org/10.1109/TCE.2022.3228404
  3. H.S.Lee, "A method for optimizing lifetime prediction of a storage device using the frequency of occurrence of defects in NAND flash memory," Journal of Internet of Things and Convergence, Vol.7, No.4, pp.9-14, 2021. https://doi.org/10.20465/KIOTS.2021.7.4.009
  4. H.S.Lee, "High Efficiency Life Prediction and Exception Processing Method of NAND Flash Memory-based Storage using Gradient Descent Method," Journal of Convergence for Information Technology, Vol.11, No.11, pp.44-50, 2021 https://doi.org/10.22156/CS4SMB.2021.11.11.044
  5. H.S.Lee, "A Safety IO Throttling Method Inducting Differential End of Life to Improving the Reliability of Big Data Maintenance in the SSD based RAID," Journal of Digital Convergence, Vol.20, No.5, pp.593-598, 2022. https://doi.org/10.14400/JDC.2022.20.5.593
  6. H.S.Lee, "Performance analysis and prediction through various over-provision on NAND flash memory based storage," Journal of Digital Convergence, Vol.20, No.3, pp.343-348, 2022. https://doi.org/10.14400/JDC.2022.20.3.343
  7. S.Yoo, and D.Shin, "Reinforcement Learning-Based SLC Cache Technique for Enhancing SSD Write Performance," In 12th USENIX Workshop on Hot Topics in Storage and File Systems, 2020.
  8. S.Y.Lim and H.J.Chang, "Airflow management analysis to suppress data center hot spots," Building and Environment, Vol.197, pp.107843, 2021.
  9. X.Ye and S.Xu, "Study on surface defect classification of hot-rolled strip based on PSO-SVM," In Proceedings of the Eighth Asia International Symposium on Mechatronics, Vol.885, pp.1846-1855, 2021.
  10. S.Rajendran, O.I.Khalaf, Y.Alotaibi and S.Alghamdi, "MapReduce-based big data classification model using feature subset selection and hyperparameter tuned deep belief network," Scientific Reports, Vol.11, No.1, pp.24138, 2021.
  11. S.Wu, W.Zhang, B.Mao and H.Jiang, "HotR: Alleviating Read/Write Interference with Hot Read Data Replication for Flash Storage," 2019 Design, Automation & Test in Europe Conference & Exhibition (DATE), pp.1367-1372, 2019.
  12. H.S.Lee, "A Prediction-Based Data Read Ahead Policy using Decision Tree for improving the performance of NAND flash memory based storage devices," Journal of Internet of Things and Convergence, Vol.8, No.4, pp.9-15, 2022. https://doi.org/10.20465/KIOTS.2022.8.4.009
  13. H.S.Lee, "A Memory Mapping Technique to Reduce Data Retrieval Cost in the Storage Consisting of Multi Memories," Journal of Internet of Things and Convergence, Vol.9, No.1, pp.19-24, 2023 https://doi.org/10.20465/KIOTS.2023.9.1.019
  14. H.S.Lee, "A Study on the Performance Measurement and Analysis on the Virtual Memory based FTL Policy through the Changing Map Data Resource," Journal of Internet of Things and Convergence, Vol.9, No.1, pp.71-76, 2023 https://doi.org/10.20465/KIOTS.2023.9.1.071
  15. http://iotta.snia.org/traces/block-io/388