【摘要】：隨著互聯(lián)網(wǎng)的發(fā)展,網(wǎng)絡(luò)用戶數(shù)量迅速增長,用戶的需求日益豐富,各種新的無線接入技術(shù)相繼出現(xiàn),并旨在為用戶帶來越來越豐富的業(yè)務(wù)體驗。網(wǎng)絡(luò)虛擬化,實現(xiàn)了底層網(wǎng)絡(luò)的數(shù)據(jù)轉(zhuǎn)發(fā)功能和控制功能的解耦與分離,并能夠根據(jù)用戶創(chuàng)建虛擬請求,動態(tài)創(chuàng)建虛擬網(wǎng)絡(luò),其本質(zhì)是:減少基站的數(shù)量,共享底層物理資源。通過虛擬化技術(shù)將底層物理資源抽象成虛擬資源,然后對虛擬資源進行橫向切片,再將切片資源按照指定的資源調(diào)度算法調(diào)度,用于承載虛擬請求。所以,如何高效管理并調(diào)度虛擬資源用于承載用戶請求是虛擬資源管理的研究重點。首先,介紹了已有的虛擬資源管理架構(gòu)的工作流程以及工作原理,并總結(jié)歸納了各架構(gòu)模型的特點。已有的虛擬資源管理架構(gòu)主要有:分布式管理架構(gòu)、集中式管理架構(gòu)、集中式管理并且分布式控制的架構(gòu)模型。然后,以光和無線融合組網(wǎng)為研究對象,研究了基于小世界模型的虛擬資源管理算法。該算法通過對底層物理網(wǎng)絡(luò)構(gòu)建小世界模型,采用虛擬化技術(shù)將底層資源抽象成虛擬資源,用于承載虛擬請求。論文針對節(jié)點間的路徑長度、數(shù)據(jù)傳輸時延和節(jié)點吞吐量進行了數(shù)值仿真。仿真結(jié)果表明,基于小世界模型的虛擬資源管理算法降低了節(jié)點間路徑長度降低和數(shù)據(jù)傳輸時延,提高了節(jié)點吞吐量。最后,本文提出了基于小世界模型和用戶優(yōu)先級的多路徑(SWPMP)映射算法。該算法定義:用于承載虛擬請求的路徑不再是一條路徑,而是多條路徑共同承載虛擬請求,每條路徑上按照指定的算法分配需要傳輸?shù)臄?shù)據(jù)量,因此,完成數(shù)據(jù)傳輸所用的時間是每條路徑上的時延的最大值。傳輸時延的降低,使得節(jié)點處的吞吐量增加。同時,結(jié)合用戶優(yōu)先級對資源管理的影響,優(yōu)先級高的用戶請求優(yōu)先被響應(yīng),而優(yōu)先級低的滯后響應(yīng),保證了不同業(yè)務(wù)類型的虛擬請求有次序的被響應(yīng)。論文針對映射過程中的數(shù)據(jù)傳輸時延和節(jié)點吞吐量進行了數(shù)值仿真,仿真中考慮了用戶請求優(yōu)先級對算法性能的影響。由仿真結(jié)果可以得出,基于小世界模型和用戶優(yōu)先級的多路徑(SWPMP)映射算法與基于小世界模型的虛擬資源管理算法相比,進一步降低了數(shù)據(jù)傳輸時延,有效均衡了網(wǎng)絡(luò)負載,并提高了節(jié)點吞吐量。
[Abstract]:With the development of the Internet, the number of network users is growing rapidly, and the demand of users is becoming more and more abundant. A variety of new wireless access technologies have emerged one after another, and the purpose is to bring more and more business experience to users. Network virtualization realizes the decoupling and separation of data forwarding function and control function of the underlying network and can create virtual network dynamically according to the user's virtual request. The essence of virtual network is to reduce the number of base stations and share the underlying physical resources. The underlying physical resources are abstracted into virtual resources by virtualization technology, then the virtual resources are sliced horizontally, and then the sliced resources are scheduled according to the specified resource scheduling algorithm to carry the virtual requests. Therefore, how to efficiently manage and schedule virtual resources to host user requests is the focus of virtual resource management. Firstly, the workflow and working principle of the existing virtual resource management architecture are introduced, and the characteristics of each architecture model are summarized. The existing virtual resource management architecture includes: distributed management architecture, centralized management and distributed control architecture. Then, the virtual resource management algorithm based on small world model is studied. The algorithm constructs a small world model for the underlying physical network and abstracts the underlying resources into virtual resources using virtualization technology to carry virtual requests. In this paper, the path length, data transmission delay and node throughput between nodes are numerically simulated. The simulation results show that the virtual resource management algorithm based on small world model reduces the path length and data transmission delay between nodes and improves the throughput of nodes. Finally, a multipath (SWPMP) mapping algorithm based on small world model and user priority is proposed. The algorithm defines that the path used to host the virtual request is no longer one path, but that multiple paths together carry the virtual request, and each path allocates the amount of data to be transferred according to the specified algorithm, so, The time taken to complete the data transfer is the maximum delay on each path. The reduction of transmission delay increases the throughput at the node. At the same time, considering the influence of user priority on resource management, the high priority user request is responded first, while the low priority delay response ensures that the virtual request of different service types is responded in order. In this paper, the data transmission delay and node throughput in the mapping process are numerically simulated, and the effect of user request priority on the performance of the algorithm is considered in the simulation. The simulation results show that the multi-path (SWPMP) mapping algorithm based on small-world model and user priority can further reduce the delay of data transmission and balance the network load effectively compared with the virtual resource management algorithm based on small-world model. The node throughput is improved.
【學位授予單位】：南京郵電大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TN915.6

【參考文獻】

相關(guān)期刊論文 前1條

1 趙慧玲;馮明;史凡;;SDN——未來網(wǎng)絡(luò)演進的重要趨勢[J];電信科學;2012年11期

，

本文編號：2303928