.png){kind=icon}
Return
Efficient task-based intermittent computing leveraging SRAM data retention
.png){kind=icon}
PRE
AI
DOI：10.1007/s11227-025-08205-7.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Energy harvesting has the potential to overcome the limitations of batteries and power numerous Internet of Things devices. However, the energy generated by the harvesters is often weak and unstable, leading to frequent and unpredictable power failures. This can result in cyclic reboots without any forward progress in the system. To address this issue, a task-based intermittent computing system is proposed that periodically saves system states into non-volatile memory (NVM). Although this approach resolves the problem of non-progress, it requires frequent backups, which introduces a new challenge of reducing backup overhead. This is a significant research problem in intermittent computing. Our work introduces an innovative solution that leverages the ability of volatile SRAM to retain data for seconds to minutes after a power loss. This capability allows systems to maintain progress even during frequent power failures. We present an energy-aware backup strategy and a hybrid memory management policy that aims to minimize NVM usage while ensuring the consistency of system states. Additionally, we have developed an efficient method to validate a subset of program states using the cyclic redundancy check (CRC), which ensures the integrity of SRAM data after a failure. Our runtime system, which incorporates these techniques, demonstrates the ability to enable continuous progress on an MSP430 platform with minimal overhead.
Keywords:
Energy harvesting
Intermittent computing system
SRAM data retention
Journal
T
IF:
0
Papers:
647
Citations:
0
Organization
Cited Papers
.png){kind=icon}
ShareSaveShareSaveHibernus plus plus : A Self-Calibrating and Adaptive System for Transiently-Powered Embedded Devices
PREAI
ShareSave
View more