Research On Resource Scheduling Strategy Of Cloud Computing Platform In Electric Power System

The data type and scale of the smart grid are increasing and accumulated at an unprecedented rate, which marking electric power system has entered the era of big data. The big data of electric power system brings a great challenge of storage, management and analysis capabilities for the grid information platform, and also the existing hardware and computing capacity has been difficult to adapt to the computing and storage capacity that required by the electric power system of on-line analysis and real-time control in the future. Cloud computing-based virtualization extensions for its flexible, powerful storage capacity and computing power, causing widespread concern domestic and foreign. With the widespread concerned domestic and foreign about the cloud computing by its flexible extensions, powerful storage capacity and computing abality, academics have done a lot of research on introducing the cloud computing tecknowledge into the electic power system of to build a unified electricity cloud computing platform. However, the construction of the cloud computing platform in electric power system will also introducing a common problem of cloud computing, such as single point of failure of cluster, quality of service issues, effective utilization of resources and energy issues.With the analysis of the research about the cloud computing platform of enecltic power system, single point of failure of cloud computing and the status of cloud computing resource scheduling in this paper. Firstly, combined with the electric power systems business development needs and the latest developments in cloud computing it gives a electric power cloud computing platform architecture based on open source Open Stack and Hadoop. Then in view of the presence of the single point of failure of the master node in Hadoop, and gives a kind of hot standby master control node to resolve the single point of failure. Finally, it gives a resource scheduling strategy based on dynamic migration of virtual machines by analyzing the resource scheduling problem about the cloud computing platform of electric power. In response to these problems above the paper gives a detailed analysis of the virtual machine migration process in the resource scheduling strategy, and in the process of the virtual machine migration the strategy determines the hotspot by the double threshold policy based on smoothness index prediction model, selects the virtual machine to migtate by the combination of migration effect and speed, at the phase of selecting the migration target, the strategy search for the target host by particle swarm optimization algorithm based on annealing thought, and use the roulette optimization to determine the target host to reach the long-term optimization effect of migration. In addition, the objectives of optimization consider three aspects including the quality of service, resource utilization and energy cost of the clusters.According to the characteristics of the simulation of electric power system, we use the Cloud Sim to simulate the cloud computing platform of power simulation with the initialization of virtual machine placement and the resource scheduling strategy. Combined with the hot spots determination based on double threshold prediction and the selection policy which considers the effects and time consummation of virtual machine migration that put forward by this paper, and then conpares the experiment result of the long-term optimization of particle swarm which based on annealing thinking, the standard PSO algorithm, the greedy algorithm with no hot spots prediction and the virtual machine sequence placement algorithm with no migration, and the comparison of experimental results verify the particle swarm optimization algorithm combined with resource scheduling policy that given in this paper has the significant optimization effect on the aspects of the service level agreement(SLA), the remaining resource utilization, the power consumption of the platform and the number of migration. And the experiment also verify that the resource scheduling strategy proposed by this paper can ensure the efficient and stable operation of the power of cloud computing platform and also provide a feasible solution to build and to optimize the resource of the cloud computing platform of electric power.