1 of 34

Achieving Energy Efficiency on the Edge Through Containerization: A Brownout Approach

2 of 34

TABLE OF CONTENTS

MOTIVATION

RESEARCH QUESTIONS

RESEARCH TIMELINE

PROBLEM STATEMENT

OBJECTIVES

LITERATURE REVIEW

METHODOLOGY

DELIVERABLES

01

03

06

02

07

04

05

08

3 of 34

Motivation

01

4 of 34

MOTIVATION

  • Edge computing is gaining popularity due to numerous applications in the current context, mainly in IoT. Some of the applications are;
    • Oil Refineries - Fume detection, maintaining pipe pressure and temperature.
    • Portable applications used by emergency response teams and military personnel.
    • Seismic wave detection applications.
    • Home Automation.

5 of 34

MOTIVATION

  • Energy efficiency in edge computing is critical due to the limited energy capacity available in edge devices. Furthermore, edge devices are usually powered by batteries and renewable energy sources, which can lead to inconsistencies in power supply.

​

  • Research has been conducted to investigate energy efficiency in edge computing environments at infrastructure level.
    • Task-offloading
    • Dynamic Resource Scheduling
    • Energy scheduling
    • DVFS

​

  • Another case where energy efficiency is considered are in large-scale cloud environments.
    • VM Consolidation
    • DVFS
    • Brownout

​

  • Brownout approach is a container orchestration based approach which involves activating and deactivating microservices or application components to reduce energy usage without affecting the main functionality of the application.

6 of 34

MOTIVATION

  • There are numerous challenges in edge computing.
    • Weak computing power.
    • Limited memory capacity.
    • Limited storage capacity.
    • Limited energy capacity.

​

  • Due to the above challenges. an approach that accomplishes the demanded application goals and minimizes energy consumption with low overhead is required.

​

  • Given the immense popularity of containerized applications in edge computing environments, we realize the need for a novel approach that utilizes a container orchestration based approach for optimizing energy consumption on the edge.

​

  • Since the brownout approach has not been explored in the edge computing domain and as it has been successful in large scale clouds, we are motivated to use the brownout-based approach for edge devices.

​

7 of 34

Problem Statement

02

8 of 34

PROBLEM STATEMENT

​

  • Edge computing has a multitude of applications but it has numerous challenges due to the availability of very limited resources and energy.

​

  • One of the major challenges for applications in edge devices is low energy availability.

​

  • Maintaining application demands such as low latency is vital while achieving energy efficiency.

​

  • Containerized applications are popular due to portability, flexibility, ease of scalability, etc.

​

  • However, there is a lack of research investigating the problem of low energy availability for containerized applications in edge computing environments.

​

  • We propose to address this problem by utilizing a container orchestration based approach for optimizing energy efficiency of applications on the edge.

9 of 34

Research Questions

03

10 of 34

RESEARCH QUESTIONS

​

​

  • How to achieve energy efficiency in resource constrained edge devices?

​

  • How to maintain QoS such as response time while reducing energy consumption in edge devices?

​

  • How does containerized applications help to reduce energy usage in edge computing environments?

​

11 of 34

Objectives

04

12 of 34

OBJECTIVES

​

​

  • To enhance the energy efficiency of container-based applications in edge devices.

​

  • To design a strategy to reduce energy consumption while maintaining QoS requirements.

​

  • To reduce the computational overhead of the energy reduction strategy.

​

13 of 34

Literature Review

05

14 of 34

Literature Review - Taxonomy

15 of 34

Literature Review

  • We classify related works in two ways according to the computation environment, and then categorize the work by approaches used to achieve energy efficiency in the respective computing environment, as shown above.

​

  • Achieving energy efficiency in Cloud.
    • Brownout approach
    • Dynamic Resource Allocation + Brownout
    • VM Consolidation
    • DVFS

​

  • Achieving energy efficiency in Edge Computing.
    • Brownout approach
    • Dynamic Resource Scheduling
    • Task offloading
    • DVFS

16 of 34

Achieving energy efficiency in cloud

​

  • Brownout approach
    • Xu et al. [4] propose a prospective model for brownout approach in cloud computing systems based on the MAPE-K [14] loop control model.
    • The same authors have proposed a prototype model based on brownout scheduling algorithms called iBrownout [5] and a software system called BrownoutCon [6].
    • It is based on scheduling algorithms that utilizes a brownout approach to optimize energy usage in microservice-based applications in large-scale clouds.
    • This shows that brownout is a viable approach to reduce energy usage at container orchestration level.

​

​

17 of 34

Achieving energy efficiency in cloud

​

​

  • Dynamic Resource Allocation + Brownout
    • Hasan et al. [7] propose GPaaScaler model, which dynamically allocates/ deallocates physical resources at infrastructure level while using a brownout approach at application level to achieve energy efficiency.
    • Beldiceanu et al, [8] introduce EpoCloud prototype, which uses a brownout-based approach at application level for web jobs considering green energy awareness, and utilizes VM consolidation with a high-speed optics network at infrastructure-level to optimize energy efficiency.

18 of 34

Achieving energy efficiency in Edge Computing

​

  • Dynamic Resource Scheduling
    • Aslanpour et al. [10]. propose a model which identifies edge nodes that face bottlenecks due to power outages, overloading and node failure, and a centralized scheduler transfers its functions to other working nodes without compromising QoS.
  • Task Offloading
    • Ullah et al. [11] propose to offload tasks from a resource-constrained device to a resource-rich device to optimize the energy consumption in edge devices.
  • Brownout approach
    • Chen et al. [13] have proposed a container selection policy which uses brownout that considers penalty and resource utilization equilibrium. However, this paper has not been published in a well recognized journal and the research conducted is unsatisfactory.

19 of 34

Summary

Research

Conference/ Journal

Cloud

Edge

Container Based

Infrastructure Level

Container Based

Infrastructure Level

Tang et al. [1]

IJGUC, 2016

​

✔

​

​

Safari et al. [2]

Simul Model Pract Theory, 2018

​

✔

​

​

Beloglazov et al. [3]

Concurr Comput,, 2011

​

✔

​

​

Xu et al. [4]

ACM Comput. Surv., 2019

✔

​

​

​

Xu et al. [5]

IEEE T-SUSC, 2019

✔

​

​

​

Xu et al. [6]

JSS, 2019

✔

​

​

​

Hasan et al. [7]

UCC 2017

✔

✔

​

​

20 of 34

Summary

Research

Conference/ Journal

Cloud

Edge

Container Based

Infrastructure Level

Container Based

Infrastructure Level

Beldiceanu et al. [8]

Computing, 2017

✔

✔

​

​

Gu et al. [9]

FGCS, 2019

​

​

​

✔

Aslanpour et al. [10]

CCGrid 2022

​

​

​

✔

Ullah et al. [11]

arXiv preprint , 2022

​

​

​

✔

Panda et al. [12]

IEEE IoT-J

​

​

​

✔

Chen et al. [13]

IHMSC 2019

​

​

✔

​

Proposed Approach

​

​

​

✔

​

21 of 34

Methodology

06

22 of 34

METHODOLOGY

23 of 34

METHODOLOGY

​

  • Microservice based application - Consists of mandatory and optional services.

​

  • There are two modules in the system
    • Analytical model
    • Brownout controller
      • Pre-processor + Brownout algorithm + Actuators

​

  • Input for the analytical model
    • Characterization of containerized application
    • Characterization of edge computing cluster
    • Power consumption reading of edge cluster (for verification purposes)

​

  • Output from the analytical model to the brownout controller
    • Computed power consumption
    • Future user demand

​

  • Output from the brownout controller
    • Activate and deactivate decision for optional containers

24 of 34

IMPLEMENTATION

​

  • The experimental setup for evaluating the proposed approach will consist of;
    • A cluster consisting of multiple ”Raspberry Pi 3 Model B Plus” [15] devices.
    • K3s container orchestration system [16].
    • A power monitor connected to the cluster to get power readings.

​

​

  • For testing purposes, we will be using a containerized application deployed on the edge cluster.
    • Eg:- Home surveillance and automation system with a machine learning model for people detection.

25 of 34

VALIDATION AND EVALUATION

​

  • A power monitor will be used in taking power and energy measurements to validate the correctness of the analytical model.
  • We will perform a controlled experiment where the application will be run with and without the brownout controller to validate the effectiveness of our approach.
  • Our approach will also be evaluated and compared against the existing techniques used for achieving energy efficiency.

26 of 34

Research Timeline

07

27 of 34

Gantt Chart

28 of 34

Deliverables

08

29 of 34

Deliverables

​

  • Research artifacts
    • Brownout algorithm
    • Analytical model to compute the power consumption
    • Research paper
  • Open source software client to control and monitor the brownout-enabled edge cluster.

​

30 of 34

REFERENCES

[1] Z. Tang, L. Qi, Z. Cheng, K. Li, S. U. Khan, and K. Li, “An Energy-Efficient Task Scheduling Algorithm in DVFS-enabled Cloud Environment,” J Grid Computing, vol. 14, no. 1, pp. 55–74, Mar. 2016, doi: 10.1007/s10723-015-9334-y.

​

[2] M. Safari and R. Khorsand, “Energy-aware scheduling algorithm for time-constrained workflow tasks in DVFS-enabled cloud environment,” Simulation Modelling Practice and Theory, vol. 87, pp. 311–326, Sep. 2018, doi: 10.1016/j.simpat.2018.07.006.

​

[3] A. Beloglazov and R. Buyya, “Optimal online deterministic algorithms and adaptive heuristics for energy and performance efficient dynamic consolidation of virtual machines in Cloud data centers: ENERGY AND PERFORMANCE EFFICIENT DYNAMIC CONSOLIDATION OF VIRTUAL MACHINES,” Concurrency Computat.: Pract. Exper., vol. 24, no. 13, pp. 1397–1420, Sep. 2012, doi: 10.1002/cpe.1867.

31 of 34

REFERENCES

[4] M. Xu and R. Buyya, “Brownout approach for adaptive management of resources

and applications in cloud computing systems: A taxonomy and future directions,”

ACM Computing Surveys (CSUR), vol. 52, no. 1, pp. 1–27, 2019.

​

[5] M. Xu, A. N. Toosi, and R. Buyya, “iBrownout: An Integrated Approach for Managing Energy and Brownout in Container-Based Clouds,” IEEE Trans. Sustain. Comput., vol. 4, no. 1, pp. 53–66, Jan. 2019, doi: 10.1109/TSUSC.2018.2808493.

​

[6] M. Xu and R. Buyya, “BrownoutCon: A software system based on brownout and containers for energy-efficient cloud computing,” J. Syst. Softw., vol. 155, pp. 91–103, Sep. 2019, doi: 10.1016/j.jss.2019.05.031.

​

[7] M. S. Hasan, F. Alvares, and T. Ledoux, “GPaaScaler: Green Energy Aware Platform Scaler for Interactive Cloud Application,” in Proceedings of the10th International Conference on Utility and Cloud Computing, Austin Texas USA, Dec. 2017, pp. 79–89. doi: 10.1145/3147213.3147227.

​

​

32 of 34

REFERENCES

[8] N. Beldiceanu et al., “Towards energy-proportional clouds partially powered by renewable energy,” Computing, vol. 99, no. 1, pp. 3–22, Jan. 2017, doi: 10.1007/s00607-016-0503-z.

​

[9] L. Gu, J. Cai, D. Zeng, Y. Zhang, H. Jin, and W. Dai, “Energy efficient task allocation and energy scheduling in green energy powered edge computing,” Future Gener. Comput. Syst., vol. 95, pp. 89–99, Jun. 2019, doi: 10.1016/j.future.2018.12.062.

​

[10] M. S. Aslanpour, A. N. Toosi, M. A. Cheema, and R. Gaire, “Energy-Aware Resource Scheduling for Serverless Edge Computing,” in 2022 22nd IEEE International Symposium on Cluster, Cloud and Internet Computing (CCGrid), Taormina, Italy, May 2022, pp. 190–199. doi: 10.1109/CCGrid54584.2022.00028.

​

[11] F. Ullah, I. Mohammed, and M. A. Babar, “A Framework for Energy-aware Evaluation of Distributed Data Processing Platforms in Edge-Cloud Environment,” ArXiv Prepr. ArXiv220101972, 2022.

​

33 of 34

REFERENCES

​

[12] S. K. Panda, M. Lin, and T. Zhou, “Energy Efficient Computation Offloading with DVFS using Deep Reinforcement Learning for Time-Critical IoT Applications in Edge Computing,” IEEE Internet Things J., pp. 1–1, 2022, doi: 10.1109/JIOT.2022.3153399.

​

[13] F. Chen and X. Zhou, “The Container Selection Policy Based on Brownout in Edge Computing,” in 2019 11th International Conference on Intelligent Human-Machine Systems and Cybernetics (IHMSC), Hangzhou, China, Aug. 2019, pp. 97–100. doi: 10.1109/IHMSC.2019.10118.

​

[14] P. Arcaini, E. Riccobene, and P. Scandurra, “Modeling and analyzing MAPE-K feedback loops for self-adaptation,” 2015, pp. 13–23.

​

[15] R. Pi, “Buy a raspberry pi 3 model b+.” [Online]. Available: https://www.raspberrypi.com/products/raspberry-pi-3-model-b-plus/

​

[16] “K3s.” [Online]. Available: https://k3s.io/

​

​

​

​

34 of 34

THANK YOU!