臺灣博碩士論文加值系統

隨著物聯網的蓬勃發展，其應用領域日漸廣泛，依據不同的需求發展成許多服務，例如智慧交通、智慧醫療、智慧農業等，相對的也提升了系統複雜性，而延伸出諸多需要解決的問題，包括(1)資料類型之異質性提高，所需的流量皆不相同，單一的傳輸方式並不適用於各種資料類型，(2)龐大的終端設備數量所發出的連線要求可能造成系統阻塞(block)，以及(3)如何妥善配置各項連線要求所需要的資源等皆為物聯網之研究議題。本論文提出一「高效能整合式物聯網中介平台」(A High Efficient Hybrid IoT Middleware，HEH IoT Middleware)，整體架構包含了(1)整合式資訊傳輸中介平台(Hybrid Information Transmission Middleware, HITM)、(2)細緻資源配置機制(Fine-grained Resource Allocation Mechanism, FRA)、(3)流量預測機制(Traffic Prediction Mechanism, TP)，以及(4)感測服務閘道器(Sensor Service Gateway, SSG)四個部分。本論文之網路架構採用叢集(Cluster)架構，在中介平台和終端設備之間設計一SSG作為雙方溝通的媒介，由SSG管理數個終端設備並負責判斷感測資料是否異常，以及向中介平台發送連線要求，以此方式減少終端設備向中介平台的連線數量。HEH IoT Middleware整合VM、Event、Agent三種系統控制類別架構，可以依據不同應用情境選擇合適的運算類別，其中由VM負責傳送占用較多資源的圖片或影片、Event負責日常或定期記錄的文字資料，而Agent則負責異常資料以及控制指令的傳送。其次，HEH IoT Middleware利用Docker輕量級虛擬化技術實作此VM、Event、Agent三種運算類別架構，並採用FRA機制將系統資源細分成數個Container，可讓各個傳送任務並行處理，進一步縮短系統的平均傳送時間。最後，為了降低系統阻塞(block)機率，本論文設計一TP機制，透過m個機器學習方法產出m個預測結果和m個誤差，再導入卡爾曼濾波器(Kalman Filter)進行m-1次的資料交正，以縮小誤差並取得更接近真實的流量預測，於系統運算效能改進上可依據此預測值分析系統阻塞機率，進而控管Container的開啟數量。

With the vigorous development of Internet of Things, its application is becoming more and more extensive. According to the different needs of the development of many services, the relative also enhance the complexity of the system, and extend a lot of problems to be solved, include (1) the heterogeneity of the data type is improved, the required traffic is not the same, a single transmission is not applicable to a variety of data types, (2) the number of large terminal equipment issued by the connection requirements may cause the system block (block), (3) how to properly configure the connection requirements of the resources required for the Internet of Things research topics. This paper introduce a High Efficient Hybrid IoT Middleware (HEH IoT Middleware), includes (1) Hybrid Information Transmission Middleware (HITM), (2) Fine-grained Resource Allocation Mechanism (FRA), (3) Traffic Prediction Mechanism (TP), and (4) Service Gateway, SSG). In this paper, a cluster architecture is used to design a SSG between the intermediary platform and the terminal equipment as a medium, via SSG management of a number of terminal equipment and is responsible for judging whether the sensing data is abnormal, and send the connection request to the intermediary platform, in this way to reduce the number of terminal equipment to the intermediary platform. HEH IoT Middleware integrates VM, Event, Agent three kinds of system control category architecture, according to different application scenarios to select the appropriate computing category, which is responsible for the transfer of VMs occupy more resources for the picture or video, Event is responsible for daily or regular recording of text data, Agent is responsible for the transmission of exception data and control instructions. Followed by the use of Docker lightweight virtualization technology to implement this VM, Event, Agent three types of computing architecture, and use the FRA mechanism to subdivide the system resources into several Containers, which can make the transmission tasks parallel processing, and further shorten the average transmission time of the system. Finally, in order to reduce the chances of the system, this paper designs a TP mechanism to produce the prediction results and errors through the machine learning method, and then introduce the Kalman Filter to correct the data and narrow the error and get closer to the real traffic forecast. The improvement can be based on this forecast value analysis system blocking probability, and then control the number of Container open.

摘要 I
Abstract III
謝誌 V
目錄 VI
圖索引 IX
表索引 XI
1 緒論 1
1.1 研究背景與動機 1
1.2 研究目的與方法 2
1.3 論文架構 4
2 技術背景與文獻探討 5
2.1 物聯網中介平台(Internet of Thing Middleware) 5
2.1.1 虛擬機器基礎(VM-based) 5
2.1.2 事件基礎(Event-based) 7
2.1.3 代理人基礎(Agent-based) 9
2.1.4 文獻比較 10
2.1.5 類別比較 11
2.2 Docker 12
2.2.1 Docker特色 13
2.2.2 Docker架構 14
2.2.3 基本概念 15
2.3 卡爾曼濾波器(Kalman Filter) 16
2.3.1 基礎模型 17
2.3.2 遞迴(recursive) 18
3 研究架構與方法 22
3.1 網路架構設計 22
3.2 研究架構設計 23
3.2.1 整合式資訊傳輸中介平台(Hybrid Information Transmission Middleware) 23
3.2.2 細緻資源配置機制(Fine-grained Resource Allocation Mechanism) 26
3.2.3 流量預測機制(Traffic Prediction Mechanism) 27
3.2.4 感測服務閘道器(Sensor Service Gateway) 30
3.3 研究流程設計 31
3.3.1 初始設定(Initial setting) 31
3.3.2 資料傳送(Data transfer phase) 32
3.3.3 調整設定(Adjustment phase) 32
4 系統實作與效能評估 34
4.1 系統實作 34
4.1.1 實驗硬體 34
4.1.2 測試工具 35
4.1.3 整合式資訊傳輸中介平台(Hybrid Information Transmission Middleware) 35
4.1.4 細緻資源配置機制(Fine-grained Resource Allocation Mechanism) 38
4.1.5 流量預測機制(Traffic Prediction Mechanism) 40
4.1.6 感測服務閘道器(Sensor Service Gateway) 43
4.1.7 使用者介面(User Interface) 43
4.2 效能評估 46
4.2.1 服務往返(Round-Trip)時間 46
4.2.2 系統資源使用率 50
4.2.3 TP機制之資料校正 52
5 結論 55
參考文獻 56
附錄一、Docker指令說明 59
作者簡介 61

[1]Baohua Yang，「《Docker — 從入門到實踐》正體中文版」，2016。
[2]C. L. Fok, G. C. Roman, and C. Lu, “Servilla: A flexible service provisioning middleware for heterogeneous sensor networks,” Sci. Comput.Programm., vol. 77, no. 6, pp. 663–684, 2012.
[3]C. M. Kirsch, M. A. Sanvido, and T. A. Henzinger, “A programmablemicrokernel for real-time systems,” in Proc. 1st ACM/USENIX Int. Conf. Virtual Execution Environ., 2005, pp. 35–45.
[4]D. Guinard, V. Trifa, S. Karnouskos, P. Spiess, and D. Savio, “Interacting with the SOA-based Internet of Things: Discovery, query, selection, and on-demand provisioning of web services,” IEEE Trans. Serv. Comput., vol. 3, no. 3, pp. 223–235, Jul. 2010.
[5]Docker overview，https://docs.docker.com/engine/docker-overview/。(Last access : 2017/06/24)
[6]Greg Welch and Gary Bishop, “An Introduction to the Kalman Filter” , July 24, 2006
[7]J. Koshy and R. Pandey, “Vm*: Synthesizing scalable runtime environments for sensor networks,” in Proc. 3rd Int. Conf. Embedded Netw.Sensor Syst. (Sensys), 2005, pp. 243–254.
[8]J. R. Silva et al., “PRISMA: A publish-subscribe and resource-oriented middleware for wireless sensor networks,” in Proc. Adv. Int. Conf. Telecommun. (AICT’14), 2014, pp. 87–97.
[9]Josh Hu，「最完整的Docker聖經 - Docker原理圖解及全環境安裝」，2015。
[10]N. Michal, K. Artem, K. Oleksiy, N. Sergiy, S. Michal, and T. Vagan, “Challenges of middleware for the Internet of Things,” in Automation Control—Theory and Practice. InTech, 2009.
[11]P. Levis and D. Culler, “Maté: A tiny virtual machine for sensor networks,” SIGARCH Comput. Archit. News, vol. 30, no. 5, Oct. 2002.
[12]P. R. Pietzuch, “Hermes: A scalable event-based middleware,” Univ.Cambridge, Comput. Lab., Tech. Rep. UCAM-CL-TR-590, Jun.2004.
[13]Panetta, K., “Gartner’s Top 10 Strategic Technology Trends for 2017”, Gartner, Inc., 2016.
[14]Porter, M.E., Heppelmann, J.E., “How Smart, Connected Products Are Transforming Competition”, Harvard Business Review, vol. 92, no. 11, 64-88, 2014.
[15]R. E. KALMAN, “A New Approach to Linear Filtering and Prediction Problems” Journal of Basic Engineering, 82 (Series D): 35-45,1960.
[16]Ramsey Faragher, “Understanding the Basis of the Kalman Filter Via a Simple and Intuitive Derivation”, IEEE SIGNAL PROCESSING MAGAZINE, pp.128-132, SEPTEMBER 2012.
[17]Razzaque, M.A., Milojevic-Jevric, M., Palade, A., Clarke, S., “Middleware for Internet of Things: A Survey”, IEEE Internet of Things Journal, vol. 3, no. 1, pp. 70-95, 2016.
[18]T. Liu and M. Martonosi, “Impala: A middleware system for managing autonomic, parallel sensor systems,” ACM SIGPLAN Notices, vol. 38, no. 10, pp. 107–118, 2003.
[19]T. Sivaharan, G. Blair, and G. Coulson, “Green: A configurable and reconfigurable publish-subscribe middleware for pervasive computing,” in On the Move to Meaningful Internet Systems 2005: CoopIS, DOA, and ODBASE. New York, NY, USA: Springer, 2005, pp. 732–749.
[20]何宗諭，「淺談輕量化的虛擬技術 - Docker容器」，http://www.cc.ntu.edu.tw/chinese/epaper/0036/20160321_3611.html。(Last access : 2017/06/24)
[21]謝宗穎、林彥宏，「持續整合與自動化測試實作」，http://docker.readbook.tw/index.html。(Last access : 2017/06/24)