Study Of Distributed Cloud Datacenter Oriented All-Optical Interconnection Network Technologies

In the background of economic and information globalization, the new cloud services represented by smart city, wise medical and precision agriculture, mobilize various IT resource through network to provide personalized, low cost and efficient service for the user, will become the mainstream model of Internet service in the future. As an important infrastructure of cloud service, the traditional large-scale single data center(DC) has some issues, such as low utilization, high response delay, huge energy consumption, and single-point failure. Meanwhile, the design of traditional intra-DC network mainly focusing on high throughput and the inter-DC network staying on point-to-point transmission as a crude pipeline, cannot meet the demand of future cloud service for flexibility and intelligence. Against that backdrop, this dissertation puts forward all-optical interconnection Floating Cloud with the features of resource distribution and service intelligence, and studies related content including resource distributed deployment, all-optical network architecture and control mechanism, heterogeneous resources integration and cross-layer scheduling, and the strategy and algorithm of resource intelligent flow based on environmental perception. The details of related works can be summarized as follows:Firstly, a distributed all-optical interconnection Floating Cloud architecture embedding metropolitan area network is proposed. To break up the traditional model of large-scale single DC + long distance transmission, this dissertation proposes a conception of Floating Cloud,where IT resources are distributed around user to meet the demand of future service for low latency, high reliability and high utilization. In order to construct Floating Cloud, we first discuss the influencing factors of DC deployment, including external environment, the requirement of ICT capacity and location et.al, to construct optimal DC deployment model with goals of lowest cost and highest safety. Then, based on the flow characteristics and communication model of cloud DC, we propose a multi-ring topology based Floating Cloud flat network with all-optical interconnection, which can support fine-grained exchange and has good flexibility, and design the corresponding optical switch. In terms of the control plane, we use centralized OpenFlow with low complexity and supporting rapid resource scheduling. Meanwhile, the related content of OpenFlow protocol is extended to support time slot switching. In the end,an experimental platform was set up to verify the effectiveness and feasibility of the proposed architecture.Secondly, a heterogeneous resource integration architecture and cross layer scheduling mechanism for Floating Cloud is proposed. In addition to the Internected network technology, Cloud service DC involves many technologies of ICT, including computing, storage and many other heterogeneous resources. For this reason, we first discuss the necessity of heterogeneous resource integration and cross layer scheduling caused by cloud services. Then we propose a hierarchical control architecture of OpenStack+OpenFlow for IT resources integration and cross layer resource scheduling according to the characteristics of Floating Cloud.Through the experiment of dynamic adjusting virtual data center (VDC),we verify the effectiveness and feasibility of the proposed control architecture. In the end, based on the above control plane, we study some corss-layer scheduling mechanisms and optimization algorithms between application and network layer, including VDC mapping, multicast communication and Coflow scheduling.Thirdly, we propose some strategies and algorithms of resource intelligent flow in different service scenarios for Floating Cloud. In order to make full use of resource mobility to improve the performance of service,we first introduce a conception of Follow-the-X based on environment perception. After then, we propose a user access frequency based hotspot-adjusting strategy, a heuristic algorithm for hotspots deployment and migration. Through numerical simulations and experiments, we verify the feasibility of the proposed algorithm and architecture. In the end, we further discuss the strategy of seamless VDC provisioning for Floating Cloud based on user trajectory prediction, and verify the effectiveness of proposed strategy through file transfer protocol (FTP) experiment.