物聯網(IoT) 裝置透過閘道器連接至網路， 並且閘道器讓被包裝成容器的物聯網分析程式能夠轉換原始的感測器資料成為更為濃縮的處理過的資料。在這個論文裡， 我們研究兩個研究問題去最大化跑在資料中心伺服器上和閘道器上的物聯網分析程式的總體服務品質(QoS)。第一個問題是根據需要上傳的原始的感測器資料， 挑選一部分的物聯網分析程式去佈建在閘道器上，用以節省所需的上傳頻寬。第二個問題是分配剩下的上傳頻寬給所有的物聯網分析程式，用以最大化總體的服務品質。我們提出了一些演算法去解決這兩個研究問題。除此之外，我們實作了一些經典的分層替換策略並且探討了他們的表現。我們已經實作了真實的平台用以測試我們提出的系統和演算法。我們的實驗結果揭示了我們提出的演算法： (i) 運用閘道器的下載頻寬和儲存空間來節省上傳頻寬的消耗， (ii) 在沒有過載網路和閘道器的情況下，取得高服務品質級別， (iii) 在低上傳頻寬的環境下，比起其他兩個基準算法，服務品質級別分別高出了18%和37%， (iv) 在高上傳頻寬的環境下，比起其他兩個基準算法，上傳頻寬的使用率分別高出
了162%和61%。

Internet-of-Things (IoT) devices are connected to the Internet through a gateway, which can host IoT analytics encapsulated in containers to convert raw sensor data into more condensed processed data. In this thesis, we study two research problems to maximize the overall Quality-of-Service (QoS) level of all IoT analytics that run on both data center servers and gateways. The first problem is selecting additional IoT analytics to deploy on a gateway to save upload bandwidth due to uploading raw sensor data. The second problem is allocating the residue upload bandwidth among all IoT analytics to maximize the overall QoS level. We propose several algorithms to solve these two research problems. Moreover, we implement several classical layer replacement policies and discuss their performance. We have implemented real testbeds to evaluate our proposed system and algorithms. Our experiment results reveal that our proposed algorithms: (i) capitalize the download bandwidth and storage space of the gateway in order to save the upload bandwidth consumption, (ii) achieve high QoS levels without overloading the network and gateway, (iii) outperform the other two baseline algorithms by 18% and 37% in QoS levels in low upload network bandwidth environment, and (iv) outperform the other two baseline algorithms by 162% and 61% in the utilization rate of upload bandwidth in high upload network bandwidth environment.

Acknowledgments i
致謝 ii
Abstract iii
中文摘要 iv
1 Introduction 1
1.1 Contributions 2
1.2 Thesis Organization 3
2 Background 4
2.1 Edge Computing 4
2.2 Internet of Things 5
2.3 IoT Analytics 6
2.3.1 Docker 7
2.3.2 Kubernetes 7
3 Research Problem 9
3.1 Problem Statement 9
3.2 Problem Decomposition 10
4 System Architecture 12
4.1 Server and Controller 12
4.2 Gateway 13
5 Image Download Problem and Algorithms 15
5.1 Problem Formulation 16
5.2 Dynamic Programming Algorithm 17
5.3 (1 − ǫ)-Approximation Algorithm 18
5.4 Greedy Algorithm 19
6 Rate Allocation Problem and Algorithms 20
6.1 QoS and Bandwidth Models of IoT Analytics 21
6.2 Problem Formulation 22
6.3 Rate Allocation Algorithm 22
6.4 Analysis 23
7 Layer Replacement Policies 25
8 Evaluations 28
8.1 Implementations 28
8.2 Testbeds 31
8.3 Setup 32
8.4 Results 35
8.4.1 Default Sample Run Analysis 35
8.4.2 Image Download Algorithm Analysis 36
8.4.3 Rate Allocation Algorithm Analysis 40
8.4.4 Layer Replacement Policy Analysis 41
9 Related Work 44
9.1 IoT Platforms without Edge Devices 44
9.2 IoT Platforms with Edge Devices 45
9.3 IoT Analytics on Edge Devices 47
10 Conclusion 48
Bibliography 50

[1] Amazon echo. https://www.amazon.com/Amazon-Echo-Bluetooth-Speaker-with-Alexa-Black/dp/B00X4WHP5E/.
[2] Audio classification: Multilayer neural networks using TensorFlow. https://github.com/nextco/audio-classification.
[3] Docker web page. https://www.docker.com/.
[4] Global IoT analytics market to grow at a CAGR of +30.9% during forecast period 2018-2025. https://www.marketresearchfuture.com/reports/iot-analytics-market-1757.
[5] Google home. https://store.google.com/gb/product/google home/.
[6] Internet of things (iot) connected devices installed base worldwide from 2015 to 2025. https://www.statista.com/statistics/471264/iot-number-of-connected-devicesworldwide/.
[7] Kubernetes web page. https://kubernetes.io/.
[8] Linux containers web page. https://linuxcontainers.org/.
[9] Moby: open framework created by docker to assemble specialized container systems. https://mobyproject.org/.
[10] Tensorflow web page. https://www.tensorflow.org/.
[11] Tmall genie. https://bot.tmall.com/.
[12] Urban sound 8k dataset. https://urbansounddataset.weebly.com/urbansound8k.html.
[13] Visual object classes challenge 2012. http://host.robots.ox.ac.uk/pascal/VOC/voc2012/#data.
[14] YOLO: Real-time object detection. https://pjreddie.com/darknet/yolo/.
[15] K. Ashton. That ’internet of things’thing. RFID journal, 22(7):97–114, 2009.
[16] L.-J. Chen, Y.-H. Ho, H.-H. Hsieh, S.-T. Huang, H.-C. Lee, and S. Mahajan. ADF: An anomaly detection framework for large-scale PM2.5 sensing systems. IEEE Internet of Things Journal, 5(2):559–570, April 2018.
[17] Z. Chu, F. Zhou, Z. Zhu, R. Q. Hu, and P. Xiao. Wireless powered sensor networks for Internet of Things: Maximum throughput and optimal power allocation. IEEE Internet of Things Journal, 5(1):310–321, February 2018.
[18] R. Deng, R. Lu, C. Lai, T. H. Luan, and H. Liang. Optimal workload allocation in fog-cloud computing toward balanced delay and power consumption. IEEE Internet of Things Journal, 3(6):1171–1181, December 2016.
[19] S. Ezdiani, I. S. Acharyya, S. Sivakumar, and A. Al-Anbuky. Wireless sensor network softwarization: Towards WSN adaptive QoS. IEEE Internet of Things Journal, 4(5):1517–1527, October 2017.
[20] Q. Fan and N. Ansari. Application aware workload allocation for edge computing-based IoT. IEEE Internet of Things Journal, 5(3):2146–2153, June 2018.
[21] M. A. A. Faruque and K. Vatanparvar. Energy management-as-a-service over fog computing platform. IEEE Internet of Things Journal, 3(2):161–169, April 2016. 

[22] J. He, J. Wei, K. Chen, Z. Tang, Y. Zhou, and Y. Zhang. Multitier fog computing with large-scale IoT data analytics for smart cities. IEEE Internet of Things Journal, 5(2):677–686, April 2018. 

[23] H. Hong, P. Tsai, A. Cheng, M. Uddin, N. Venkatasubramanian, and C. Hsu. Sup- porting internet-of-things analytics in a fog computing platform. In IEEE Interna- tional Conference on Cloud Computing Technology and Science (CloudCom), Hong Kong, China, 2017. 

[24] H. Hong, P. Tsai, and C. Hsu. Dynamic module deployment in a fog computing platform. In in Proc. of Asia-Pacific Network Operations and Management Sympo- sium (APNOMS), Kanazawa, Japan, 2016. 

[25] A.Howard,M.Zhu,B.Chen,D.Kalenichenko,W.Wang,T.Weyand,M.Andreetto, and H. Adam. Mobilenets: Efficient convolutional neural networks for mobile vision applications. In arXiv preprint arXiv:1704.04861., 2017. 

[26] J. Huang, Y. Meng, X. Gong, Y. Liu, and Q. Duan. A novel deployment scheme for green Internet of Things. IEEE Internet of Things Journal, 1(2):196–205, April 2014. 

[27] Z. Ji, I. Ganchev, O. Droma, L. Zhao, and X. Zhang. A cloud-based car park- ing middleware for iot-based smart cities: Design and implementation. Sensors, 14(12):22372–22393, November 2014. 

[28] D. Katsaros and Y. Manolopoulos. Cache management for web-powered databases. In J. W. Rahayu and D. Taniar, editors, Web-Powered Databases, chapter 8, pages 203–244. IGI Global, Hershey, PA, USA, 2003. 

[29] H. Kellerer and U. Pferschy. Knapsack Problems. Springer, 2004. 

[30] A. Kiani and N. Ansari. Toward hierarchical mobile edge computing: An auction- based profit maximization approach. IEEE Internet of Things Journal, 4(6):2082– 2091, December 2017.
[31] B. Li and J. Yu. Research and application on the smart home based on component technologies and internet of things. Procedia Engineering, 15:2087–2092, 2011. CEIS 2011. 

[32] L. Liu, Z. Chang, X. Guo, S. Mao, and T. Ristaniemi. Multiobjective optimization for computation offloading in fog computing. IEEE Internet of Things Journal, 5(1):283–294, February 2018. 

[33] K. Lv, J. Hu, Q. Yu, and K. Yang. Throughput maximization and fairness assurance in data and energy integrated communication networks. IEEE Internet of Things Journal, 5(2):636–644, April 2018. 

[34] M. H. Y. Moghaddam and A. Leon-Garcia. A fog-based internet of energy architec- ture for transactive energy management systems. IEEE Internet of Things Journal, 5(2):1055–1069, April 2018. 

[35] R. Morabito, I.Farris, A. Iera, and T. Taleb. Evaluating performance of containerized IoT services for clustered devices at the network edge. IEEE Internet of Things Journal, 4(4):1019–1030, August 2017. 

[36] B. Mostafa, A. Benslimane, M. Saleh, S. Kassem, and M. Molnar. An energy- efficient multiobjective scheduling model for monitoring in Internet of Things. IEEE Internet of Things Journal, 5(3):1727–1738, June 2018.
[37] Openfog reference architecture for fog computing. Standard, OpenFog Consortium, 2017. 

[38] C. Pahl, S. Helmer, L. Miori, J. Sanin, and B. Lee. A container-based edge cloud paas architecture based on raspberry pi clusters. In Proc. of IEEE International Conference on Future Internet of Things and Cloud Workshops (FiCloudW’16), page 117–124, Vienna, Austria, October 2016.

[39] C. Pahl and B. Lee. Containers and clusters for edge cloud architectures–a technol- ogy review. In Proc. of IEEE International Conference on Future Internet of Things and Cloud Workshops (FiCloudW), Rome, Italy, August 2015. 

[40] D. Patterson and J. Hennessy. Computer Organization and Design: the Hardware Software Interface. Morgan Kaufmann, 2004.
[41] M. Rahman, A. Rahman, H. J. Hong, L. W. Pan, M. Y. S. Uddin, N. Venkatasubra- manian, and C. H. Hsu. An adaptive iot platform on budgeted 3g data plans. Journal of Systems Architecture, 97:65–76, August 2019.
[42] P. M. Santos, J. Rodrigues, S. Cruz, T. Lourenc ̧o, P. d’Orey, Y. Luis, C. Rocha, S. Sousa, S. Criso ́stomo, C. Queiro ́s, S. Sargento, A. Aguiar, and J. Barros. Porto- livinglab: An iot-based sensing platform for smart cities. IEEE Internet of Things Journal, 5(2):523–532, April 2018.
[43] S. Verma, Y. Kawamoto, Z. Fadlullah, H. Nishiyama, and N. Kato. A survey on network methodologies for real-time analytics of massive iot data and open research issues. IEEE Communications Surveys & Tutorials, 19(3):1457–1477, April 2017.
[44] P. Wang, C. Yao, Z. Zheng, G. Sun, and L. Song. Joint task assignment, trans- mission, and computing resource allocation in multilayer mobile edge computing systems. IEEE Internet of Things Journal, 6(2):2872–2884, April 2019.
[45] Y. Wei, F. Yu, M. Song, and Z. Han. Joint optimization of caching, computing, and radio resources for fog-enabled iot using natural actor–critic deep reinforcement learning. IEEE Internet of Things Journal, 6(2):2061–2073, April 2019.
[46] Y. Xu and A. Helal. Scalable cloud–sensor architecture for the Internet of Things. IEEE Internet of Things Journal, 3(3):285–298, June 2016.
[47] A. Yousefpour, G. Ishigaki, R. Gour, and J. P. Jue. On reducing IoT service delay via fog offloading. IEEE Internet of Things Journal, 5(2):998–1010, April 2018.
[48] W. Yu, F. Liang, X. He, W. G. Hatcher, C. Lu, J. Lin, and X. Yang. A survey on the edge computing for the internet of things. IEEE Access, 6:6900–6919, 2018. 

[49] A. Zanella, N. Bui, A. Castellani, L. Vangelista, and M. Zorzi. Internet of Things for smart cities. IEEE Internet of Things Journal, 1(1):22–32, 2014. 

[50] D. Zhai, R. Zhang, L. Cai, B. Li, and Y. Jiang. Energy-efficient user scheduling and power allocation for NOMA-based wireless networks with massive IoT devices. IEEE Internet of Things Journal, 5(3):1857–1868, June 2018. 

[51] X. Zhai, X. Guan, C. Zhu, L. Shu, and J. Yuan. Optimization algorithms for multiac- cess green communications in Internet of Things. IEEE Internet of Things Journal, 5(3):1739–1748, June 2018. 

[52] L. Zhao, W. Sun, Y. Shi, and J. Liu. Optimal placement of cloudlets for access delay minimization in (SDN-based) Internet of Things networks. IEEE Internet of Things Journal, 5(2):1334–1344, April 2018.