Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Optimizing Embedded Software of Self-Powered IoT Edge Devices for Transient Computing

Pan, Chen (2020) Optimizing Embedded Software of Self-Powered IoT Edge Devices for Transient Computing. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

[img]
Preview
PDF
Download (4MB) | Preview

Abstract

IoT edge computing becomes increasingly popular as it can mitigate the burden of cloud servers significantly by offloading tasks from the cloud to the edge which contains the majority of IoT devices. Currently, there are trillions of edge devices all over the world, and this number keeps increasing. A vast amount of edge devices work under power-constrained scenarios such as for outdoor environmental monitoring. Considering the cost and sustainability, in the long run, self-powering through energy harvesting technology is preferred for these IoT edge devices. Nevertheless, a common and critical drawback of self-powered IoT edge devices is that their runtime states in volatile memory such as SRAM will be lost during the power outage. Thanks to the state-of-the-art non-volatile processor (NVP), the runtime volatile states can be saved into the on-chip non-volatile memory before the power outage and recovered when harvesting power becomes available. Yet the potential of a self-powered IoT edge device is still hindered by the intrinsic low energy efficiency and reliability.

In order to fully exert the potentials of existing self-powered IoT edge devices, this dissertation aims at optimizing the energy efficiency and reliability of self-powered IoT edge devices through several software approaches. First, to prevent execution progress loss during the power outage, NVP-aware task schedulers are proposed to maximize the overall task execution progress especially for the atomic tasks of which the unfinished progress is subjected to loss regardless of having been checkpointed. Second, to minimize both the time and energy overheads of checkpointing operations on non-volatile memory, an intelligent checkpointing scheme is proposed which can not only ensure a successful checkpointing but also predict the necessity of conducting checkpointing to avoid excessive checkpointing overhead. Third, to avoid inappropriate runtime CPU clock frequency with low energy utility, a CPU frequency modulator is proposed which adjusts the runtime CPU clock frequency adaptively. Finally, to thrive in ultra-low harvesting power scenarios, a light-weight software paradigm is proposed to help maximize the energy extraction rate of the energy harvester and power regulator bundle. Besides, checkpointing is also optimized for more energy-efficient and light-weight operation.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Pan, Chenchen.pan@pitt.educhp115
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairHu, Jingtongjthu@pitt.edu
Committee CoChairMao, Zhi-Hongmaozh@engr.pitt.edu
Committee MemberDickerson, Samueldickerson@pitt.edu
Committee MemberXiong, Fengf.xiong@pitt.edu
Committee MemberZhang, Youtaozhangyt@cs.pitt.edu
Committee MemberJones, Alexakjones@pitt.edu
Date: 29 January 2020
Date Type: Publication
Defense Date: 30 May 2019
Approval Date: 29 January 2020
Submission Date: 9 November 2019
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 115
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Electrical and Computer Engineering
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: IoT, Edge Computing, Embedded Systems, Energy Harvesting, Non-volatile Memory, Non-volatile Processor
Date Deposited: 29 Jan 2020 16:23
Last Modified: 29 Jan 2020 16:23
URI: http://d-scholarship.pitt.edu/id/eprint/37765

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item