近年來，物聯網（Internet-of-Things）應用程式所產生的大量數據皆需要強大的分析方法，例如深度學習，來提取出有用的資訊，而現有的物聯網應用程式大多將數據傳輸到計算資源豐富的數據中心來進行分析。但是，大量的數據可能會造成網絡的壅塞、數據中心的嚴重負擔、以及增加安全上的漏洞。在我的論文中，我設計了一個採用霧計算概念的平台，將數據中心（服務器）和終端運算裝置（物聯網運算裝置）的資源進行整合。它有兩個特點：（i）動態部署和（ii）邊緣分析。我在位於邊端的運算裝置之間執行分佈式的分析應用程式來對數據進行預處理，而不是將所有資料都完整地發送到數據中心。然後，我分析了實作這樣一個平台所必須面對的難題，並採用擁有龐大社群支持的開源專案來克服這些挑戰。最後，我對實作出的的平台進行全面的測試，結果顯示了：（i）分佈式分析的好處及其局限性，（ii）當跨多個運算裝置分發應用程式時，工作分配的重要性，以及（iii）平台中所使用的工具所帶來的額外運算成本。

Modern Internet-of-Things (IoT) applications produce a large amount of data and require powerful analytics approaches, such as using Deep Learning to extract useful information. Existing IoT applications transmit the data to resource-rich data centers for analytics. However, it may congest networks, overload data centers, and increase security vulnerability. In my thesis, I implement a platform, which adopts the concept of Fog Computing, integrating resources from data centers (servers) to end devices (IoT devices). It has two features: (i)dynamic deployment and, (ii) edge analytics. I launch distributed analytics applications among the devices to pre-process the data, rather than sending everything to the data centers. I analyze the challenges to implement such a platform and carefully adopt popular open-source projects to overcome the challenges. I then conduct comprehensive experiments on the implemented platform. The results show: (i) the benefits/limitations of distributed analytics, (ii) the importance of decisions on distributing an application across multiple devices, and (iii) the overhead caused by different components in my platform.

致謝 i
Acknowledgments ii
中文摘要 iii
Abstract iv
1 Introduction 1
1.1 Challenges.................................. 2
1.2 Contributions ................................ 3
1.3 ThesisOrganization............................. 4
2 Background 5
2.1 InternetofThings.............................. 5
2.2 FogComputing ............................... 6
2.3 EdgeAnalytics ............................... 7
3 Dynamic Deployment 9
3.1 Container-BasedVirtualization....................... 10
3.2 OrchestrationTool ............................. 11
3.3 Event-DrivenMechanism.......................... 13
4 Edge Analytics 15
4.1 Location-based and Sensor-based Services for the Internet of Things . . . 16
4.2 DistributedComputing ........................... 16
4.3 Pre-processingwithDeepLearning..................... 17
5 System Overview 20
5.1 Master.................................... 20
5.2 Minions................................... 23
6 Implementation and Demo Scenarios 24
6.1 Implementation ............................... 24
6.1.1 TensorFlow............................. 24
6.1.2 Docker ............................... 25
6.1.3 Kubernetes ............................. 25
6.2 Scenarios .................................. 26
6.2.1 ObjectDetection .......................... 27
6.2.2 SoundClassification ........................ 28
6.2.3 AirQualityMonitor ........................ 29
7 Evaluations 31
7.1 Setup .................................... 31
7.2 Results.................................... 32
8 Related Work 39
8.1 FogComputingPlatform .......................... 39
8.2 EdgeAnalyticsintheInternetofThings.................. 40
9 Conclusion and Future Work 41
Bibliography 42

[1] Amazon Echo. https://www.amazon.com/Amazon-Echo-Bluetooth-Speaker-with- WiFi-Alexa/dp/B00X4WHP5E.
[2] Augmented Traffic Control. http://facebook.github.io/augmented-traffic-control/.
[3] CloudPing.Info. http://www.cloudping.info.
[4] Docker. https://www.docker.com.
[5] Gartner says 8.4 billion connected ”things” will be in use in 2017, up 31 percent from 2016. http://www.gartner.com/newsroom/id/3598917.
[6] Gartner says a typical family home could contain more than 500 smart devices by 2022. http://www.gartner.com/newsroom/id/2839717.
[7] HypriotOS. https://blog.hypriot.com.
[8] iperf. https://iperf.fr.
[9] Kubernetes. http://kubernetes.io/.
[10] KVM. http://www.linux-kvm.org/.
[11] librosa. https://github.com/librosa/librosa.
[12] LXC. https://linuxcontainers.org.
[13] MQTT. http://mqtt.org.
[14] Opencv. http://opencv.org.
[15] OpenStack. https://www.openstack.org/.
[16] Saltstack. https://saltstack.com/.
[17] Swarm. https://docs.docker.com/swarm/overview/.
[18] TensorFlow. https://www.tensorflow.org.
[19] Urbansound8kdataset.https://serv.cusp.nyu.edu/projects/urbansounddataset/urbansound8k.html.
[20] Wonder Shaper. http://lartc.org/wondershaper/.
[21] Wonder Shaper. https://linux.die.net/man/8/netstat.
[22] Xen. http://www.xenproject.org/.
[23] Fog computing and the Internet of Things: Extend the cloud to where the things are. http://www.cisco.com/c/dam/en_us/solutions/ trends/iot/docs/computing-overview.pdf, 2015.
[24] K. Ashton. That ’internet of things’ thing. RFID Journal, pages 97–114, 2009.
[25] L. Atzori, A. Iera, and G. Morabito. The internet of things: A survey. Computer
networks, 54(15):2787–2805, 2010.
[26] F. Bonomi, R. Milito, J. Zhu, and S. Addepalli. Fog computing and its role in the internet of things. In Proc. of ACM Workshop on Mobile Cloud Computing (MCC), 2012.
[27] R. Friedman, A. Kogan, and Y. Krivolapov. On power and throughput tradeoffs of WiFi and Bluetooth in smartphones. IEEE Transactions on Mobile Computing, 12(2):1363–1375, 2013.
[28] K. Giang, M. Blackstock, R. Lea, and C. Leung. Developing IoT applications in the fog: A distributed dataflow approach. In Proc of IEEE International Conference on Internet of Things (IOT), 2015.
[29] N. Giang, M. Blackstock, R. Lea, and V. Leung. Developing IoT applications in the fog: A distributed dataflow approach. In Proc. of IoT, 2015.
[30] H. Hong, J. Chuang, and C. Hsu. Animation rendering on multimedia fog com- puting platforms. In Proc. of IEEE International Conference on Cloud Computing Technology and Science (CloudCom), Luxembourg, December 2016.
[31] 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 Cloud Comput- ing Technology and Science (CloudCom), 2017.
[32] H. Hong, P. Tsai, and C. Hsu. Dynamic module deployment in a fog computing platform. In Proc. of Asia-Pacific Network Operations and Management Symposium (APNOMS), 2016.
[33] K. Hong, D. Lillethun, U. Ramachandran, O. B, and B. Koldehofe. Mobile fog: a programming model for large-scale applications on the internet of things. In Proc. of ACM Workshop on Mobile Cloud Computing (MCC), 2013.
[34] K. Hong, D. Lillethun, U. Ramachandran, B. Ottenwalder, and B. Koldehofe. Mo- bile fog: a programming model for large-scale applications on the internet of things. In Proc. of ACM SIGCOMM workshop on Mobile Cloud Computing (MCC), 2013.
[35] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi. You only look once: Unified, real-time object detection. In Proc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 779–788, 2016.
[36] M. Satyanarayanan, P. Bahl, R. Ca ́ceres, and N. Davies. The case for VM- based cloudlets in mobile computing. IEEE Transactions on Pervasive Computing, 8(4):14–24, 2009.
[37] E.Saurez,K.Hong,D.Lillethun,U.Ramachandran,andB.Ottenwa ̈lder.Incremen- tal deployment and migration of geo-distributed situation awareness applications in the fog. In Proc. of ACM International Conference on Distributed and Event-based Systems (DEBS), 2016.
[38] S. Shin, S. Seo, S. Eom, J. Jung, and H. Lee. A Pub/Sub-Based fog computing architecture for Internet-of-Vehicles. In Proc. of IEEE International Conference on Cloud Computing Technology and Science (CloudCom), 2016.
[39] M. Y. S. Uddin, A. Nelson, K. Benson, G. Wang, Q. Zhu, Q. Han, N. Alhassoun, P. Chakravarthi, J. Stamatakis, D. Hoffman, L. Darcy, and N. Venkatasubrama- nian. The SCALE2 multi-network architecture for iot-based resilient communities. In Proc. of IEEE International Conference on Smart Computing (SMARTCOMP), pages 1–8, May 2016.
[40] I.T.Union.Theinternetofthings—executivesummary.ITUInternetReports,2005.
[41] L. Vaquero and L. Merino. Finding your way in the fog: Towards a comprehensive definition of fog computing. ACM SIGCOMM Computer Communication Review, 44(5):27–32, 2014.
[42] D. Willis, A. Dasgupta, and S. Banerjee. ParaDrop: a multi-tenant platform for dynamically installed third party services on home gateways. In Proc. of ACM SIG- COMM Workshop on Distributed Cloud Computing (DCC), 2014.
[43] D.Wu,I.Arkhipov,M.Kim,L.Talcott,C.Regan,A.McCann,andV.N.ADDSEN: adaptive data processing and dissemination for drone swarms in urban sensing. IEEE Transactions on Computers, 66(2):183–198, 2016.
[44] S. Yi, Z. Hao, and Q. Li. Fog computing: Platform and applications. In Proc. of IEEE Workshop on Hot Topics in Web Systems and Technologies (HotWeb), 2015.