Pseudo-Log: Restore Global Data Facing Power Failures with Minimum NVM Write

Edwin Sha, Yeteng Liao, Qingfeng Zhuge, Rui Xu, Yuhong Song, Jialin Liu

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Embedded devices and systems are commonly used in various scenarios. But they often face power failures because of the unstable power supply. So it is important to back up the running state and global data of the program under this architecture. Meanwhile, non-volatile memory (NVM) is widely used to back up the program state and store global data because of its byte-addressability, low access latency, and persistency. But NVM has a limited write endurance and the backup procedure may cause many writes to NVM. There are many works to optimize the backup of the program state, but few considers the consistency between the program state and global data. In this paper, we propose a scheme, Pseudo-Log, to delay the update of global data and then reduce the write on NVM and guarantee the global data is restored correctly. Experiments show that our method greatly reduces the number of NVM writes compared with other methods. The average improvement in the number of writes is about 160x.

Original languageEnglish
Title of host publicationProceedings - 24th IEEE International Conference on High Performance Computing and Communications, 8th IEEE International Conference on Data Science and Systems, 20th IEEE International Conference on Smart City and 8th IEEE International Conference on Dependability in Sensor, Cloud and Big Data Systems and Application, HPCC/DSS/SmartCity/DependSys 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages2027-2030
Number of pages4
ISBN (Electronic)9798350319934
DOIs
StatePublished - 2022
Event24th IEEE International Conference on High Performance Computing and Communications, 8th IEEE International Conference on Data Science and Systems, 20th IEEE International Conference on Smart City and 8th IEEE International Conference on Dependability in Sensor, Cloud and Big Data Systems and Application, HPCC/DSS/SmartCity/DependSys 2022 - Chengdu, China
Duration: 18 Dec 202220 Dec 2022

Publication series

NameProceedings - 24th IEEE International Conference on High Performance Computing and Communications, 8th IEEE International Conference on Data Science and Systems, 20th IEEE International Conference on Smart City and 8th IEEE International Conference on Dependability in Sensor, Cloud and Big Data Systems and Application, HPCC/DSS/SmartCity/DependSys 2022

Conference

Conference24th IEEE International Conference on High Performance Computing and Communications, 8th IEEE International Conference on Data Science and Systems, 20th IEEE International Conference on Smart City and 8th IEEE International Conference on Dependability in Sensor, Cloud and Big Data Systems and Application, HPCC/DSS/SmartCity/DependSys 2022
Country/TerritoryChina
CityChengdu
Period18/12/2220/12/22

Keywords

  • backup
  • delay update
  • global data
  • power failure
  • stack frame

Fingerprint

Dive into the research topics of 'Pseudo-Log: Restore Global Data Facing Power Failures with Minimum NVM Write'. Together they form a unique fingerprint.

Cite this