| With the development of cloud computing, bigdata and other new businesses, centralized data center has become an important infrastructure to support cloud computing and the development of bigdata, thus, constructing centralized data center to meet the cloud computing and big data development is imminent. Meanwhile, the amount of data for storage and processing of large enterprises showed a geometric growth trend, constructing centralized data centers has become one of the most important works at present.Maintenance of cloud computing and deployment of big data, which cause dataexplosive growth and the number of servers exponential increasing. What’s more, data transfer mode between centralized data center become more and more complex, Business requirements on QoS become more and more high, and the demand of fault self-healing becomes more and more strong. Based on traditional distributed control mode, according to the user’s service request, data center make allocation of computing and storage resources, and then through the physical network interface to request the transmission channel. According to the requirement of the size of the business, time delay and so on, physical network utilizes routing protocol and control signaling to make routing calculation, bandwidth allocation and reservation, and then to complete the data transmission. However, data center and physical network are lacking of unified scheduling mechanism, so only through specific interface can make service requests, and it is difficult to achieve the global optimization of resource efficiency. When the network is congested, the physical network will reject all businessrequest of data center, networkcannot according to the service level prior to manage the transmission of high level services, and cannot guarantee business QoS. On the other hand, in recent years, the frequent occurrence of natural disasters, thelocal war weapons of mass destruction, as well as law breakers destruction are all serious threat to a centralized data center content security, so to keep the data center connectivity in the data center breakdown is urgent. Therefore, the efficiency of resources, the sensitivity of the business and the relevance of content are difficult technical problems that the current data center networkingfacing.Software defined network (SDN) can realize the control and forwarding separation, support the application of open programmable interface. SDN will pull out the core of the network function and control function from the physical device, to achieve the devices centralized management through a centralized controller and a communication protocol. This paper utilizes the advantages of SDN, and SDN will be applied to the centralized data center network, we put forward a software defined network integrated control architecture in centralized data center optical interconnection scenarios, abstract thedata center application resources and the transmitting cyber source, what’s more, through the unified control scheduling strategies to achieve global optimization. On this basis, this paper make a thorough researchbased on the network survivability of contents connectivity, cross layer optimization design of service awareness and cyber source redirection processing and other aspects. Besides, we build the software defined centralized data center network test platform (SD-CDCN) to verifythe key technology performance and make data analysis. The main innovative works of this paper are shown as follows.(1) Content connectivity theory and control technology on the disaster sceneThe establishment of a centralized data center provides the new protection scheme for the protection of network node failure on disaster scenarios. Protection theory from the original network connectivity is extended to the contents connectivity, the point-to-point service request change to point-to-either data center resources businessrequest. This paper analyzes the current centralized data center network, establishes a disaster scene model, and introduces content connectivity principle. Based on the content connectivity, we put forward the strategies of bandwidth compression and energy optimization, design the ILP model and protection-restoration strategies, what’s more, we make simulation in electric power communication network topology.(2)Resources cross layer optimization technology based on service perceptionAccording to the business needs of centralized data center and resource optimization of business oriented adaptation problem, this paper analyzes the business demands and characteristics of centralized data center, proposes a cross layer optimization technology based on centralized data center application resources and cyber source, designs a feature oriented routing, modulation format and spectrum allocationalgorithm. Meanwhile, Service aware resource optimization performance of the algorithm is verified by experiment simulation platform, which improve the performance of the whole network blocking rate of centralized data center network.(3) Resource redirection technology based on the sharing componentsFor pressing problems of centralized data center network such as rigidity and inefficiency, taking software defined network SDN technology as the core architecture, based on the physical layer technology such as the data plane and network components, we propose the redirect infrastructure resources theory, technology and algorithm. Westudy the basic functional unit of the centralized data center network, develop the sharing technology based on hardware components. Facing the functional requirements of Centralized data center network, we set up network function unit model, and design a method on a common platform to realize network virtualization, and then improve theapplication efficiency of network resources. |
PDF Full Text Request