Research On Intelligent Routing And Management Collaborative Mechanism For Cloud Computing At Priority Of Energy Efficiency

Cloud computing as a new computing service model, is connected through the Internet, using the open technologies and standards to make the hardware and software resources as the dynamic resources, and provided to users in the form of services. Data center as a foundation relying facilities of cloud computing platform, is expanding with unprecedented scale under the driven of cloud computing technology. However, the high energy consumption, low resource utilization, environmental pollution and other issues of data center, have been restricting the development of the data center for a long time. Cloud computing makes storage and computing capability to migrate to remote resources, such as virtual services and storage systems are mostly hosted in the data center (DC). This migration can bring significant energy saving; if effectively use local resources, the greenhouse gas emissions of ICT (Information and Communication Technology) will be reduced by 40%. Thus, cloud service providers need to handle the energy consumption brought by transmission networks and data centers carefully.In this thesis, we re-examine the issue of energy consumption in the context of cloud computing, study energy efficiency routing and management of cloud service; and we would comprehensively consider the energy consumption of data centers and transport network, re-enacted energy consumption model. The network which supports cloud computing needs to transfer large amounts of data in a fast and reliable way; given optical network based on wavelength division multiplexing (WDM) technology has the performance of high data rate and low latency, so select optical network as a transport network is very appropriate. Therefore, in this thesis we focus on reducing the energy consumption of integrated optical network and the IT infrastructure.First of all, we introduce the concept of cloud computing and data center; against the shortcomings of the traditional data center, we focus on introducing the characteristics of cloud computing data center; and analyze the green energy-saving problem of data center network in detail. In order to solve the energy problem of the transmission network, we introduce the IP over WDM network and discuss the energy-saving strategies of IP over WDM network.Secondly, based on the cloud-specific anycast principle we intelligently select the appropriate data center for the users and then route the requests, to improve the energy efficiency of cloud services. For the energy consumption characteristics of the data center and IP over WDM network, we study intelligent routing and management collaborative mechanism at priority of energy efficiency from a centralized point, and establish a minimum energy MILP model, which aimed at minimize the total energy consumption by turning off the unused resources of data centers and transmission network. Due to the computing complexity of the large-scale network for solving the MILP model, in this thesis we put forward the optical bypass heuristic algorithm for intelligent routing based on the evolutionary game theory. In this algorithm, we not only consider to minimize the use of IP routers which consume a large amount of energy in virtual links, but also consider to close the idle servers in the data center, to minimize the total energy consumption of the transmission and processing for the services.Finally, we study the distributed joint design method of energy efficient routing and management for cloud computing, and propose energy efficient ant colony algorithm. We promptly learn the current changing network status through the use of pheromones deposited by artificial ants real-time, and then build distributed, flexible, real-time energy-efficient routing and management collaborative mechanism for cloud computing network; we also joint the network performance of control plane and traffic-aware of data plane to build energy-efficient intelligent collaborative algorithm for cloud computing network. The algorithm does not use any supervision, just let the incoming traffic flow according the accumulated pheromone and traffic center principle. Then the incoming traffic automatically aggregated on specific links, and the unused links gradually appear through the flows. By turning off the equipment which on the unused links can achieve the purpose of energy saving. At the same time, in order to facilitate the management, we also adopt the router-card sleeping strategy; and to further saving energy we study the modes of allocating router ports on router cards to the optical channels on different virtual links.Through comparative analysis of simulation experiments, the two algorithms proposed in this thesis can efficiently solve the energy-efficient routing and management issues for cloud computing, and they also show advantages at the aspects of improving the overall energy efficiency of network and IT resources, reducing energy waste and improving quality of cloud services.