Skip to Main Content
 

Global Search Box

 
 
 

ETD Abstract Container

Abstract Header

Building Energy-efficient Edge Systems

Tumkur Ramesh Babu, Naveen
http://orcid.org/0000-0002-7081-3619
http://rave.ohiolink.edu/etdc/view?acc_num=osu160857422125719

Abstract Details

2020, Master of Science, Ohio State University, Computer Science and Engineering.
Autonomous systems such as self driving cars, smart traffic lights, smart homes and smart cameras are increasingly being deployed. Such systems deployed at the edge make use of low-powered edge devices and machine learning techniques in order to process inferences faster. However, such AI inference consumes precious energy, drains batteries and shortens IoT lifetimes. A variety of factors such as AI algorithm used, hardware resources available, internet capacity, latency and number of applications determine the amount of energy consumed at the edge. When deploying an autonomous application at edge, careful selection of these factors is very crucial. Firstly, we built a fully autonomous aerial system (FAAS) which uses a model-driven approach to manage edge resources in order to complete missions with feasible energy consumption. Edge resources can affect where FAAS fly and which data they sense. Usage profiles can diverge greatly across edge management policies. Our model run on FAAS benchmarks predicted throughput with 4\% error across mission, software and hardware settings. We found that model-driven management for FAAS can boost mission throughput by 10X and reduce costs by 87\%. Secondly, we explore AI-driven IoT which uses AI inference to characterize data harvested from IoT sensors. AI inference consumes precious energy, drains batteries and shortens IoT lifetimes. Given a workload with alternating inference and idle time periods, we explore scheduling techniques to perform AI inference model updates in an energy efficient way. We implemented traditional scheduling techniques such as First-come-first-served (FCFS), shortest-job-first (SJF) and least-recently-used (LRU) to observe the scheduling pattern and energy footprint required to perform updates. We use random walks to explore the space of scheduling policies and 2$^kr$ design of experiments to quantify primary effects and interactions between factors. The best random-walk policy uses much less energy than 99th and 95th percentiles. First-come-first-serve and shortest-job-first policies use 7X more energy than the best policy. In reality, an Online scheduler with a guaranteed robustness and consistency would be required to perform these updates efficiently. We extend this work by exploring an Online switching algorithm for data-driven configuration management. Lastly, we spent time on design of course project for data management in cloud. The project gives foundational hands-on experience to using relational database, Hadoop, Hive, MapReduce and implementing efficient query processing system.
Christopher Stewart (Advisor)
Yang Wang (Committee Member)
75 p.
Computer Science
Machine Learning, Internet of things, Software System, Autonomous System, Artificial Intelligence, scheduling, Random Walk, Computer Vision, Performance modeling, Edge Computing, AI-Driven IoT, FAAS, Online learning

Recommended Citations

Citations

  • Tumkur Ramesh Babu, N. (2020). Building Energy-efficient Edge Systems [Master's thesis, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu160857422125719

    APA Style (7th edition)

  • Tumkur Ramesh Babu, Naveen. Building Energy-efficient Edge Systems. 2020. Ohio State University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu160857422125719.

    MLA Style (8th edition)

  • Tumkur Ramesh Babu, Naveen. "Building Energy-efficient Edge Systems." Master's thesis, Ohio State University, 2020. http://rave.ohiolink.edu/etdc/view?acc_num=osu160857422125719

    Chicago Manual of Style (17th edition)

osu160857422125719
207
© 2020, some rights reserved.Creative Commons License Building Energy-efficient Edge Systems by Naveen Tumkur Ramesh Babu is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License. Based on a work at etd.ohiolink.edu.
This open access ETD is published by The Ohio State University and OhioLINK.