Loading...

An Adaptive and Energy Aware Data Allocation for Scratchpad Memory in Energy Harvesting Systems with Non-volatile Memory

Paridari, Ali | 2022

349 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55634 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Ejlali, Alireza
  7. Abstract:
  8. Energy consumption is considered as one of the key constraints in the design of Embedded Systems. Exploiting abundant ambient energy offers a practical solution for designing embedded systems. In this regard, emerging trends are surging towards replacing battery-powered embedded systems in various different applications such as wearable systems, Internet of Things, and pervasive computing. Also, with the development and ommercialization of emerging Non-volatile memory technologies,the potential to reduce the static energy consumption and the ability to retain data in the absence of input energy, has become practically feasible for Embedded Systems. Consequently, recent researches have presented different s olutions f or employing hybrid Scratchpad Memories, consisting of SRAM and Non-volatile Memory parts, for designing battery-powered embedded systems, aiming to improve energy consumption and/or performance. However, the Scratchpad Memory allocation and backup/restore mechanism presented in the literature, are not perfectly suitable for energy harvesting systems, due to the unpredictability of the ambient input energy and lack of guarantees for forward progress of execution. This body of research compares data allocation methods intended for embedded systems and considers the shortcomings of applying them to Energy Harvesting Systems, and therefore presents an Adaptive Data Allocation that improves energy consumption, performance and correct forward execution of EHSs, as a runtime solution. Data allocation algorithms of this research equipped with the proposed backup/restore mechanism, show considerable improvement in the adaptability of the flow of forward execution, to the changes of the amount of energy stored in the system capacitor. Simulations pertaining to the proposed method of this research, demonstrate a significant increase in the number of cycles of correct execution (12.65% and 17.47%), along with a considerable decrease in the number of simulation cycles spent on backup/restore of memory and charging of the system capacitor by the harvester (16.46% and 22.40%), comparing to two of the prominent related works in the literature
  9. Keywords:
  10. Energy Harvesting ; Energy Consumption ; Nonvolatile Memory ; Adaptive Data Allocation ; Energy Reduction ; Embedded System Design

 Digital Object List