雲端運算為現今網路服務的焦點領域，其透過網際網路的串連提供運算服務，達到軟硬體資源共享與虛擬化的資源管理，為無可取代的技術優勢與服務型態。
雲端運算服務模式可分為三種：軟體即服務（SaaS，Software as a Service）、平台即服務（PaaS，Platform as a Service）、基礎建設即服務（IaaS，Infrastructure as a Service）。本論文基於平台即服務的雲端運算平台，參考傳統虛擬機器的佈署問題與其研究，探討其中的資源佈署問題，進而設計一款雲端運算服務的排程系統。
實作上使用Docker容器和Swarm叢集管理系統作為技術底層，並以實作雲端運算系統- DDMS為基礎，改良其中資源佈署之方法。除了提升本論文的佈署目標，最大服務數，也加入即時性佈署模式的考慮。希望讓即時性服務需求擁有足夠的優先權與硬體資源，一般性的服務也能滿足基本運行需求。
本篇論文中，將資源佈屬問題以裝箱問題的演算法解決。除了探討多種裝箱問題 (Bin-packing) 演算法，並提出我們的改良設計，使其擁有批次處理 (Bulk Arrivals) 的特性，達到 Online Bin-Packing 的處理效率與 Offline Bin-Packing 的運算結果。另外，結合實作系統-DDMS，與Linux Container (LXC) 資源共享的特性，提供不同的優先權給即時性服務的容器，讓雲端運算平台的服務排程方法，也能夠解決不同服務特性的佈署問題。
最後，除了將改良演算法實作於Swarm系統，我們也以實驗來驗證不同演算法對於雲端服務系統最大服務數的績效影響。實驗模型以排隊系統代入，比較四種不同演算法的績效，並分析各參數值的關係。希望能以開發商的角度，提供符合叢集佈署需求的雲端服務平台的排程系統。

Nowadays cloud-computing is the hottest topics in the study of Internet. Through internet connection, it provided calculation service and allowed both software/hardware resources sharing and virtualization resource-management. This service has provided brand-new service style and irreplaceable advantages.
There are three types of cloud-computing service model: SaaS (Software as a Service), PaaS (Platform as a Service), IaaS (Infrastructure as a Service). Based on PaaS and researches of traditional virtual machine, the essay mainly discussed about issues of resource deployment and further designed a scheduling system with priority scheduling.
Practically, Docker Container and Swarm Cluster Management system are used as the technique basis and implementation of this essay are based on Distributed Docker Management System (DDMS). The aim of the essay is to improve its way for resource deployment: maximum the service numbers and conclude real-time factor in scheduling system. The new design is to authorize enough priority and hardware resources to the urgent request, and also satisfied the need of basic calculation service.
In the essay, we used Bin-Packing algorithm to solve resource-deployment problems. We not only discussed about different Bin-Packing algorithm but also provided the refinement method to equip with Bulk Arrivals functionality while at the same time achieving approximate efficiency of Online Bin-Packing and similar result of Offline Bin-Packing.
Finally, trying to testify each algorithm’s impact on cloud service system’s performance, we used queuing system to compared four different algorithms’ impact in our model and analyze each reference data and correlations to provide a scheduling system which aligned with the need of Cluster Deployment from developer’s viewpoint.

論文審定書 + i
謝辭 + ii
摘要 + iii
Abstract + iv
目錄 + v
圖次 + vii
表次 + ix
第一章 序論 + 10
1.1 研究動機 + 10
1.2 研究目的 + 11
1.3 論文架構 + 12
第二章 背景與相關研究 + 13
2.1 研究背景 + 13
2.1.1 雲端運算服務 + 13
2.1.2 Linux Container (LXC) + 16
2.1.3 Docker + 18
2.1.4 Swarm + 21
2.1.5 Real-Time Service + 23
2.2 相關研究 + 25
2.2.1 資源排程 + 25
2.2.2 裝箱問題 + 27
第三章 系統架構 + 33
3.1 雲端服務系統 + 33
3.2 Swarm資源排程 + 36
3.2.2 Filter + 37
3.2.3 Strategy + 38
第四章 資源佈署演算法 + 42
4.1 雲端服務平台排程器 + 42
4.2 排程設計 + 44
4.2.1 批次排程設計 + 44
4.2.2 延遲佈署設計 + 47
4.3 即時性佈署設計 + 49
4.3.1 排程器的即時性設計 + 49
4.3.2 Real-Time Containers + 50
4.3.3 容器資源的過量使用 + 51
第五章 實驗成果與效能分析 + 52
5.1 實驗設置與參數定義 + 52
5.1.1 實驗模型 + 52
5.1.2 實驗參數 + 53
5.1.3 實驗績效 + 54
5.1.4 實驗實作 + 54
5.2 最大服務數實驗 + 55
5.2.1 實驗目的 + 55
5.2.2 核心演算法比較 + 55
5.2.3 參數設置 + 56
5.2.4 實驗結果 + 57
第六章 結論與未來展望 + 67
參考文獻 + 68
附錄A + 72

[1] Peter Mell, Timothy Grance, The NIST Definition of Cloud Computing, 9/2011.
[2] LXC -Linux Containers, 6/2014, https://linuxcontainers.org/.
[3] Docker -Build, Ship, and Run Any App, Anywhere, 6/2014, http://www.docker.com/.
[4] Swarm -a Docker-native clustering system, 12/2014, https://docs.docker.com/swarm/.
[5] Digital Equipment Corporation. Digital UNIX Guide to Realtime Programming, 1996.
[6] Hadoop, Fair Scheduler. http://hadoop.apache.org/common/docs/r0.20.2/fair_scheduler.html.
[7] Hadoop, Capacity Scheduler. http://hadoop.apache.org/common/docs/r0.20.2/capacity_scheduler.html.
[8] Hindman B, Konwinski A, Zaharia M, et al. Mesos: a platform for fine-grained resource sharing in the data centers. USENIX Symposium on Networked Systems Design and Implementation. Boston, USA, 2011.
[9] Zaharia M, Borthakur D, Sen Sarma J, et al. Delay scheduling: a simple technique for achieving locality and fairness in cluster scheduling. EuroSys Conference. Paris, France, 2010.
[10] A. Bestavros, T. Cheatham, Jr., and D. Stefanescu. Parallel Bin packing using first fit and k-delayed best-fit heuristics. In Parallel and Distributed Processing, 1990. Proceedings of the Second IEEE Symposium on, 1990, pp. 501- 504.
[11] D. S. Johnson. Near-Optimal Bin Packing Algorithms. Ph.D. thesis, Department of Mathematics, Massachusetts Institute of Technology, Cambridge, MA, 1973.
[12] Richard E. Korf. A New Algorithm for Optimal Bin Packing. In Proceedings of the Eighteenth National Conference on Artificial Intelligence and Fourteenth Conference on Innovative Applications of Artificial Intelligence, pages 731–736. AAAI/IAAI, 2002.
[13] Yujia Ge, Guiyi Wei. GA-Based Task Scheduler for the Cloud Computing Systems. IEEE, 2010
[14] O.R. Kelly, H. Aydin, and B. Zhao. On partitioned scheduling of fixed priority mixed-criticality task sets. In Proc. of the 8th IEEE International Conference on Embedded Software and Systems (ICESS), 2011.
[15] Andr´as Gy¨orgy, G´abor Lugosi, Gy¨orgy Ottucs´ak. On-Line Sequential Bin Packing. Journal of Machine Learning Research, vol 11, 2010.
[16] Richard J. Anderson, Ernst W. Mayr, Manfred K. Warmuth. Parallel Approximation Algorithm for Bin Packing. Information and Computation, vol 82, no. 3, 1989.
[17] Rajdeep Dua, A Reddy Raja, Dharmesh Kakadia. Virtualization vs Containerization to support PaaS. IEEE International Conference on Cloud Engineering, 2014.
[18] M. Malathi. Cloud Computing Concepts. IEEE, 2011
[19] Paul, M., Sanyal, G. Survey and analysis of optimal scheduling strategies in cloud environment. IEEE, 2012
[20] Jeyarani, R., Ram, R. Vasanth, Nagaveni, N. Design and Implementation of an Efficient Two-Level Scheduler for Cloud Computing Environment. IEEE, 2010
[21] Huang Qi-yi, Huang Ting-lei. An optimistic job scheduling strategy based on QoS for Cloud Computing. IEEE, 2010
[22] Meng Xu, Lizhen Cui, Haiyang Wang, Yanbing Bi. A Multiple QoS Constrained Scheduling Strategy of Multiple Workflows for Cloud Computing. IEEE, 2009
[23] Hao Li, Huixi Li. A Research of Resource Scheduling Strategy for Cloud Computing Based on Pareto Optimality M×N Production Model. IEEE, 2011
[24] Jiayin Li, Meikang Qiu, Jianwei Niu, Wenzhong Gao, Ziliang Zong, Xiao Qin. Feedback Dynamic Algorithms for Preemptable Job Scheduling in Cloud Systems. IEEE, 2010
[25] Qi Cao, Zhi-Bo Wei, Wen-Mao Gong. An Optimized Algorithm for Task Scheduling Based on Activity Based Costing in Cloud Computing. IEEE, 2009
[26] Zinnen, A., Engel, T. Deadline constrained scheduling in hybrid clouds with Gaussian processes. IEEE, 2011
[27] Suresh, A., Vijayakarthick, P. Improving scheduling of backfill algorithms using balanced spiral method for cloud metascheduler. IEEE, 2011
[28] Selvarani, S., Sadhasivam, G.S. Improved costbased algorithm for task scheduling in cloud computing. IEEE, 2011
[29] Mehdi, N.A., Mamat, A.; Amer, A., Abdul-Mehdi, Z.T. Minimum Completion Time for Power-Aware Scheduling in Cloud Computing. IEEE, 2012
[30] Luna Mingyi Zhang, Keqin Li, Yan-Qing Zhang. Green Task Scheduling Algorithms with Speeds Optimization on Heterogeneous Cloud Servers. IEEE, 2011
[31] Celaya, J., Arronategui, U. A Highly Scalable Decentralized Scheduler of Tasks with Deadlines. IEEE, 2011
[32] Laiping Zhao, Yizhi Ren, Sakurai, K. A Resource Minimizing Scheduling Algorithm with Ensuring the Deadline and Reliability in Heterogeneous Systems. IEEE, 2011