- 學位論文
以智慧型多目標基因演算法設計通用多協定標籤交換路由之最佳化方法
An Optimization Method Design Generalized Multi-Protocol Label Switching Routing With Intelligent Multipurpose Genetic Algorithms.
摘要
通用多協定標籤交換GMPLS（Generalized Multi-Protocol Label Switching）是一個新型態的高速網路通訊技術，此種技術可在現今不同的網路通訊硬體及光通訊中擔任轉換、介面的角色，使不同的網路硬體系統可相互連結，共存於同一網路系統之中。光技術的進步促使光纖網路在都會網路中的需求急劇上升。目前，數據業務面臨的最為突出的挑戰就是網際網路協議(IP)與光學層的整合問題，它將成為網際網路下一階段擴展的中心環節。GMPLS大幅的改進了現在所使用的IP基底網路(IP-Based Networks)的困境，其中包含改善資訊流量工程(Traffic Engineering)，可以在網域內自動產生IP 通道（Create IP Tunnels），並可以同時跨越ATM的第二層傳輸與IP的第三層傳輸平台，簡化了網路的管理並降低系統的複雜性。但GMPLS所使用的最短路徑優先（Open Shortest Path First.OSPF）技術，採行的是Dijkstra's 演算法(Dijkstra's Algorithm)，以此得到網路傳輸所需之最短路徑樹(Shortest Path Tree)，並依此傳遞封包。但是，這種方式不足以處裡網路的實際動態條件與情況，這將會使路徑的選擇處於並非最佳狀態，也因而造成網路的部份壅塞現象。本論文針對GMPLS路由演算法進行探討，提出以基因演算法(Genetic Algorithms.GAs)之架構，以取代現行的OSPF法，基因演算法具有基因之特質，有自我演化與進化的能力，應用於GMPLS路由及訊務管理以及在路由選擇的演算法上，提供了極佳的QoS機制，並使得GMPLS傳輸的線路選擇能力，得到積極性與進步性的提昇，也使得高速光纖網路得到充份的發揮與實質的效果。
並列摘要
As a cutting-edge high-speed network communication technique, GMPLS (Generalized Multi-Protocol Label Switching) plays a role of conversion and interface in various network communication hardware and optical communications, allowing different network hardware systems to be connected with one another and co-exist in a single network system. The progress in optical techniques facilitates the skyrocketing demands for optical networks in urban network. At present, the most significant challenge in digital services is the integration of Internet protocol (IP) and optical layers, which will be the core for the next expansion of Internet. GMPLS has greatly improved the difficulties of IP-based networks used currently. The traffic engineering improvement in it automatically creates IP tunnels, as well as crosses the transmission platform of the 2nd layer transmission of ATM and that of the 3rd layer transmission of IP, simplifying the management of networks and reducing system complexity. However, the Open Shortest Path First (OSPF) technique used in GMPLS adopts the Dijkstra's Algorithm to obtain the Shortest Path Tree needed for network communications and to transmit packets. Yet, this way is unable to handle the actual dynamic conditions and situations in network, which does not place the path at the optimal status, as well as results in partial network congestion. This dissertation investigates the algorithms for GMPLS routing and proposes Genetic Algorithms (GAs) architecture to replace to current OSPF algorithm. The application of Genetic Algorithms, with their genetic characteristics and the ability of self-evolution, is not only applied to GMPLS routing, traffic management and routing algorithm, but also provides better QoS (Quality of Service), allowing improvement of aggressiveness and progressiveness in the ability of GMPLS transmission in path selection, while enabling high-speed optical networks to provide the optimized performance。
參考文獻
延伸閱讀
- Weng, J. H. (2008). 以多重解析度分析為基礎的可調式智慧型移動偵測演算法 [master's thesis, Yuan Ze University]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0009-3007200813103100
- 江倚瑄（2016）。應用多目標基因演算法於測力計拓樸最佳化〔碩士論文，國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342/NTU201601766
- Hsieh, Y. J. (2009). 多信道無線網狀網路下近似最佳化之分散式具服務品質路由演算法 [master's thesis, National Taiwan University]. Airiti Library. https://doi.org/10.6342/NTU.2009.01297
- Zhao, L., & Li, Y. (2014). Global Exponential Stability of Weighted Pseudo-Almost Periodic Solutions of Neutral Type High-Order Hopfield Neural Networks with Distributed Delays. Abstract and Applied Analysis, 2014(), 720-736-772. https://doi.org/10.1155/2014/506256
- Li, Z. (2013). Existence of Multiple Positive Periodic Solutions to Two Species Parasitical Model with Impulsive Effects and Harvesting Terms. Discrete Dynamics in Nature and Society, 2013(), 346-354-039. https://doi.org/10.1155/2013/198927