Study On Load Balancing Strategy Of Water Quality Forecast System In Jiaozhou Bay

Studying the changing process of various substances in marine water,is conducive to make predictions for the marine water environment.It is important for better governing the marine water quality environment and ensuring the sustainable development of coastal areas.Numerical simulation is mostly used in the study.In order to make the results more accurate,it is often necessary to select a vast area of computing and a rich body of water,and to improve the spatial resolution of the research area.But with the abundance of marine environmental data and the improvement of spatial resolution,the amount of computation of marine water quality prediction system has also increased dramatically,resulting that the numerical simulation process takes a long time,which makes the calculation inefficient and has some limitations.In recent years,with the development of high performance computing technology,parallelization of numerical simulation method has been applied more and more widely through MPI technology.To a certain extent,the running time of the system is saved and the performance of the system is improved.Because there is a serious imbalance between the computing tasks among nodes,the effect of reducing the running time by adding computing nodes is not obvious.More nodes can increase the message transfer,but increase the running time and reduce the efficiency of the system.In view of the above problems,this paper combines the load balancing technology with the MPI parallel technology to optimize the performance of the water quality forecast system in Jiaozhou Bay.The main research results are as follows:(1)The overall structure of the Jiaozhou bay water quality prediction system is analyzed in detail,and the distribution map of the calculation tasks in the grid is drawn,and the factors causing the load imbalance are determined,so as to find the starting point of load balancing optimization.(2)On the basis of the study of the water quality prediction system and load balancing technology and related theories in the Jiaozhou Bay,the static load balancing technology is applied to the optimization of the prediction system,and the experiments are carried out under the conditions of 2,4 and 8 nodes respectively.The load balancing effect is quantized and analyzed.The acceleration ratio and the acceleration efficiency are obviously improved.Under the 2 node condition,themaximum parallel efficiency can reach 91.098%,and the maximum parallel efficiency can reach 77.979% in the case of 4 nodes.(3)Considering the variety and complexity of the hardware environment of the system,the load information of each node is collected and processed by using the open source characteristics of the Linux system,and the additional overhead is evaluated,thus the load task between nodes is processed in time and the load between nodes is dynamically adjusted to achieve equilibrium.The dynamic load balancing of Jiaozhou bay water quality forecasting system is optimized to make the load of each node be adjusted in time,thus maintaining a relatively balanced state.Combining the advantages of static and dynamic load balancing technology,the hybrid load balancing technology is applied to the numerical simulation calculation method to improve the performance of the prediction system.2,4 and 8 nodes are tested respectively.With the 4 node as an example,the optimal operation time of the static load is higher than that of the static load,and the operation efficiency is obviously improved.The experimental results show that the optimization of the water quality forecast parallel system can effectively reduce the prediction time and improve the performance of the system by using load balancing technology.Taking 4 nodes as an example,compared with the system before load balancing optimization,the speedup of static load balancing and mixed load balancing can reach 3.1192 and 3.3070,and the highest parallel efficiency is 77.979% and 82.675%,respectively.In these two ways,the running time of the program is shortened to a certain extent,and the performance of the system has been significantly improved,but it has a better acceleration effect under the mixed load balancing mode.Therefore,the system is optimized by a hybrid load balancing model to improve the performance of the water quality forecasting system.