透過您的圖書館登入
IP:54.166.223.204
  • 學位論文

數據中心網路之基於基因演算法的能源考量流量排程

Energy Aware Flow Scheduling for Data Center Network Using Genetic Algorithms

指導教授 : 田伯隆

摘要


雲端運算(cloud computing)是近年來蓬勃發展的技術之一,而作為支撐整個雲端架構的核心,數據中心(data center)的建置也十分受重視。數據中心必須支援大量的運算以及資料的儲存與傳輸,其中順暢的網路是必不可少的,而要如何選擇網路的路徑使得我們可以達到高吞吐量(throughput)及低延遲等目標並非一個簡單的問題。同時數據中心為了保證服務,必須耗費龐大的能源,導致了高昂的操作及維護成本,並產生對環境有傷害的碳排放量。所以除了維持原本數據中心該有的效能之外,還必須考慮如何減少數據中心的耗電量。 本論文中針對近年來耗電量大量成長的數據中心的網路部分來做改善。由於交換器(Switch)的開關與否是影響網路部分耗電量的主要原因,在不需要的時候關閉交換器可以有效的降低耗電量,且通過數據中心的流量是不斷變動的,所以我們可以透過良好的路由控制來改善網路的耗電量。 在短時間內快速決定路由是一個很複雜的問題,我們在本論文中選擇利用基因演算法(Genetic Algorithm)來達到此目標,基因演算法是模仿物競天擇的法則來求解最佳化問題,屬於heuristic algorithm的一種,具有快速的特性,代價是無法保證正確性跟精確度。而本論文中也對基因演算法做出針對欲解問題的改良,以期提高求得之解的正確性。

並列摘要


Cloud computing is one of the growing technology in recent years. Data center as the core of the supporting of the entire cloud services, the topic of how to build a data center is very important. Data center should support a large number of computing and the storage and transmission of data, which needs unblocked network. However, it’s not a simple question to choose a path of the Internet which can achieve targets such as high throughput and low delay. At the same time, data centers consume huge amounts of energy to ensure performance, which causes high operational costs, and huge carbon footprints are unfriendly to the environment. Therefore, we have to consider how to reduce energy consumption and keep high performance. This thesis focus on network equipments in the data center which have rapidly growth of energy consumption recent years. The switches contribute the largest propotion of energy consumption of network equipments, so turn off unneeded switches reduce energy consumption effectively. We can develop good routing algorithm to improve energy consumption of network equipments. It’s a complicated problem to decide routing path in a short period, so we choose genetic Genetic Algorithm to achieve our goals. Genetic algorithm is one of a heuristic algorithm. It solves the optimization problem quickly by imitating the way of the natural selection. We use fat-tree topology in our simulation, and make some improvements of GA in order to fit our problem and raise the correctness of its solution.

參考文獻


[4] U.S. Environmental Protection Agency. Report to Congress on Server and Data Center Energy Efficiency. http://tinyurl.com/2jz3ft
[7] C. E. Leiserson, “Fat-Trees: Universal Networks for Hardware-Efficient Supercomputing,” IEEE Transactions on Computers, vol. 34, no. 10, October 1985, pp. 892-901.
[8] M. Al-Fares, A. Loukissas, and A. Vahdat. A Scalable, Commodity Data Center Network Architecture. In ACM SIGCOMM, pages 63-74, 2008.
[9] Holland, J. H., Adaptation in Natural and Artificial Systems, University of Michigan Press, Ann Arbor, 1975.
[12] D. Wicker, M. M. Rizki, and L. A. Tamburino, “The multi-tiered tournament selection for evolutionary neural network synthesis,” Proc. Int. Conf. Combinations of Evolutionary Computation and Neural Networks, pp.207-215, 2000.

被引用紀錄


陳宥臻(2015)。基因演算法和差分演算法在逐步移除型一區間設限資料上之可靠度評估應用〔碩士論文,淡江大學〕。華藝線上圖書館。https://doi.org/10.6846%2fTKU.2015.00231

延伸閱讀