Optimized Scheduling Method Based On Maximum Flow Model For Complex City Scene Data

Along with the development of Chinese brand new smart city,massive heterogeneous multi-modal spatio-temporal data will be generated at all times to provide data support for the visualization task of city display,analysis and exploration.City scene data is a kind of multi-modal spatio-temporal data driven by visualization tasks to satisfy specific analysis and visualization task for smart city a,which concludes but not limited to video,image,text,sound,trajectory,POI,map,real-time 3D model.With complex characteristics of multi-source,multi-granularity and multi-modal,it needs to be organized and scheduled at higher efficiency in the increasingly ubiquitous network environment.Existed research on scene data scheduling mainly focus on data preloading,prefetching,multi-thread processing and multi-level cache collaboration,which may lead to hardware resource waste and cannot fully satisfy the different scheduling requirements of multi-level tasks.With the gradual migration of GIS data centers from localization to the cloud,scene data scheduling in the cloud environment plays a key role in real-time visualization.How to improve scheduling performance to break through the network I/O bottleneck and meet the different needs of multi-level visualization tasks,turns into a severe challenge of scene data scheduling methods research.In this paper,an optimized scheduling method of complex city scene data based on maximum flow model is proposed for multi-level visualization tasks;Firstly,this paper studies the topology network of data scheduling transmission in cloud environment,and designs scheduling architecture for complex city scenes in cloud environment;Then,the maximum flow model is introduced to construct the scene data scheduling model based on the maximum flow.On this basis,the solution of complex city scene data flow is given;Further,according to the change of data access hotspots,a task driven scene data flow allocation method is developed to realize the global optimization of scene data flow in the cloud environment with limited hardware resources.The specific research contents are as follows:(1)Efficient scheduling architecture for complex city scene data in cloud environment.By analyzing the characteristics of complex city scene data and multi-level visualization task data preference and on the basis of data storage and scheduling structure in cloud environment,this paper designs a complex city scene scheduling architecture,provides a theoretical basis for scheduling optimization and through makes a break through to traditional scene data scheduling mechanism.(2)Scene data scheduling method based on maximum flow model.In this paper,the topology network of data resource scheduling is mapped to the maximum flow model,and the maximum flow scheduling model of scene data is constructed to quantify the service capability of scene data.Two task driven scene data flow adjustment mechanisms are designed,which can flexibly control the size of multi-type scene data flow and improve the service ability of scene data under limited resources.(3)The prototype system of scene data maximum flow scheduling is designed and analyzed.Through the design of scene data scheduling experiment corresponding to three levels of visualization tasks,compared with two typical methods: first come first service(FCFS)and priority scheduling algorithm(PSA),and then to the scheduling concurrency performance test of the system.The experimental results show that the maximum flow scheduling method is more flexible and efficient in adjusting the data flow of various scenes on demand,and meets the high-performance scheduling of complex city scene data with multi-task concurrency,which realizes the comprehensive optimization of scene data flow under the limited hardware resources in cloud environment.