Evaluation of Various Workloads in Filebench Suitable for Phase-change Memory

다국어 초록 (Multilingual Abstract)

Phase-Change Memory (PCM) is known as the next generation of memory thanks to its outstanding properties, such as fast speed, non-volatility, scalability, and low power consumption. Based on these characteristics, PCM can be used in larger expanded memory or faster storage compared to HDD or NAND flash memory. Therefore, various companies are trying to deploy PCM-based memory products. However, studies on deciding the target system of PCM are still insufficient, which is an obstacle to the commercialization of PCM. In this paper, a file system benchmark, Filebench, is evaluated with various operating options to find the most appropriate workload for PCM as storage. An experiment was conducted in a virtual system by mounting PCM with a PCM-aware file system. The results demonstrate that the PCM-based system performs up to 500 times better than traditional storage if executing workloads with a significant amount of write operations and synchronization operations. A number of applications were tested on various configurations of systems, and workload characteristics suitable for the PCM-based system are presented.

목차 (Table of Contents)

• Abstract
• 1. Introduction
• 2. Background
• 3. System Description
• 4. Evaluation
• 5. Conclusion
• References

참고문헌 (Reference)

1 M. Dong, "Soft updates made simple and fast on non-volatile memory" 719-731, 2017

2 M. Kim, "Segmented Tag Cache : A Novel Cache Organization for Reducing Dynamic Read Energy" 68 (68): 1546-1552, 2019

3 C. Wu, "Rethinking computer architectures and software systems for phase-change memory" 12 (12): 1-40, 2016

4 F. Bellard, "QEMU, a fast and portable dynamic translator" 41 : 46-, 2005

5 B. Kim, "PCM: Precision-Controlled Memory System for Energy Efficient Deep Neural Network Training" 1199-1204, 2020

6 J. Xu, "NOVA : A log-structured file system for hybrid volatile/non-volatile main memories" 323-338, 2016

7 "NDCTL"

8 L. Jiang, "Mitigating write disturbance in super-dense phase change memories" 216-227, 2014

9 "LIBNVDIMM: Non Volatile Devices"

10 H. Lee, "Integration and Boost of a Read-Modify-Write Module in Phase Change Memory System" 68 (68): 1772-1784, 2019

11 N. U. Mustafa, "Implications of non-volatile memory as primary storage for database management systems" 164-171, 2016

12 S. Kim, "HAD-TWL : Hot Address Detection-Based Wear Leveling for phase-change memory systems with low latency" 18 (18): 107-110, 2019

13 V. Tarasov, "Filebench : A flexible framework for file system benchmarking" 41 (41): 6-12, 2016

14 C. Lee, "Effective Parallelization of a High-Order Graph Matching Algorithm for GPU Execution" 29 (29): 560-571, 2019

15 R. Wang, "Decongest : Accelerating super-dense PCM under write disturbance by hot page remapping" 16 (16): 107-110, 2017

16 Hyokeun Lee, "Bit-width Reduction in Write Counters for Wear Leveling in a Phase-change Memory System" 대한전자공학회 9 (9): 413-419, 2020

17 M. Kim, "An OnDemand Scrubbing Solution for Read Disturbance Error in Phase-Change Memory" 110-111, 2020

18 D. T. Nguyen, "An Approximate Memory Architecture for Energy Saving in Deep Learning Applications" 67 (67): 1588-1601, 2020

19 J. Ou, "A high performance file system for non-volatile main memory" (12) : 1-16, 2016

20 L. Liang, "A case for virtualizing persistent memory" 126-140, 2016