Xiao Renzhi, Feng Dan, Hu Yuchong, Zhang Xiaoyi, Cheng Liangfeng. A Survey of Data Consistency Research for Non-Volatile Memory[J]. Journal of Computer Research and Development, 2020, 57(1): 85-101. DOI: 10.7544/issn1000-1239.2020.20190062
Citation:
Xiao Renzhi, Feng Dan, Hu Yuchong, Zhang Xiaoyi, Cheng Liangfeng. A Survey of Data Consistency Research for Non-Volatile Memory[J]. Journal of Computer Research and Development, 2020, 57(1): 85-101. DOI: 10.7544/issn1000-1239.2020.20190062
Xiao Renzhi, Feng Dan, Hu Yuchong, Zhang Xiaoyi, Cheng Liangfeng. A Survey of Data Consistency Research for Non-Volatile Memory[J]. Journal of Computer Research and Development, 2020, 57(1): 85-101. DOI: 10.7544/issn1000-1239.2020.20190062
Citation:
Xiao Renzhi, Feng Dan, Hu Yuchong, Zhang Xiaoyi, Cheng Liangfeng. A Survey of Data Consistency Research for Non-Volatile Memory[J]. Journal of Computer Research and Development, 2020, 57(1): 85-101. DOI: 10.7544/issn1000-1239.2020.20190062
1(Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan 430074)
2(School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074)
3(Shenzhen Research Institute of Huazhong University of Science and Technology, Shenzhen, Guangdong 518061)
Funds: This work was supported by the National Key Research and Development Program of China (2018YFB1003305); the National Natural Science Foundation of China (61772222) and the Shenzhen Knowledge Innovation Program (JCYJ20170307172447622).
As DRAM technology is facing the bottleneck in density scaling and the problem of high power leakage, novel non-volatile memory (NVM) has drawn extensive attention from academia and industry, due to its superiority in non-volatility, high-density, byte addressability, and low static power consumption. Novel non-volatile memory such as phase change memory (PCM) is likely to substitute or complement DRAM as system main memory. However, due to the non-volatility of NVM, when system failed, data stored in NVM may be inconsistent by reason of partial updates or memory controller write reordering. In order to guarantee the consistency of data in NVM, it is essential to ensure the serialization and persistence in NVM write operations. NVM has inherent drawbacks, such as limited write endurance and high write latency, thus reducing the number of writes can help prolong the lifetime of NVM and improve the performance of NVM-based systems as long as data consistency in NVM is guaranteed. This paper focuses on data consistency based on NVM, especially on persistent indexes, file systems and persistent transactions, and to provide better solutions or ideas for achieving low data consistency overhead. Finally, the possible research directions of data consistency based on NVM are pointed out.