Data communication optimization method and system based on regional ocean mode

Abstract

The invention relates to a data communication optimization method and system based on a regional ocean mode, and belongs to the technical field of data communication, and the method comprises the steps: obtaining an ocean mode variable in the regional ocean mode, and defining a data read-in decomposition descriptor and a data read-out decomposition descriptor of the ocean mode variable through a parallel input / output system; configuring a data read-in mapping relation and a data read-out mapping relation between the storage position of the ocean mode variable data in the storage file and the storage position in the memory for the data read-in decomposition descriptor and the data read-out decomposition descriptor respectively; the method comprises the steps of reading calculation data according to a data read-in decomposition descriptor, calculating the calculation data through a four-dimensional variational assimilation system model, outputting the calculated data according to a data read-out decomposition descriptor, and realizing efficient data acquisition and processing of ocean mode variables in a regional ocean mode by defining the data read-in decomposition descriptor and the data read-out decomposition descriptor. Data communication is optimized, the overhead of inter-process data transmission is reduced, and the calculation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data communication, and in particular to a data communication optimization method and system based on a regional ocean model. Background Art

[0002] The regional ocean model is a high-performance computing tool used for ocean numerical simulation, which can simulate the physical, chemical and biological processes of the ocean. In order to improve the accuracy of simulation and forecasting ability, the incremental four-dimensional variational assimilation system is introduced into the regional ocean model. Although the four-dimensional variational assimilation system has broad application prospects in the regional ocean model, in the ultra-large-scale computing scenario, the four-dimensional variational assimilation system still faces some technical bottlenecks, such as: The regional ocean model adopts a horizontal domain decomposition parallel strategy, and each subdomain needs to exchange boundary data with the eight-neighborhood process through a message passing interface. In the four-dimensional variational assimilation system, the nested loops (outer loop and inner loop) of forward integration, adjoint mode and tangent linear process cause the number of communications to increase in the order of O(N²), which seriously affects the computational efficiency.

[0003] As the scale of processes expands, the number of boundaries of two-dimensional domain decomposition increases with the number of processes, point-to-point communication links compete for physical bandwidth, and network congestion and delay amplification effects are significant. In a high-performance computing cluster environment, traditional communication modes are difficult to maintain linear expansion efficiency, which limits the application of four-dimensional variational assimilation systems in large-scale computing.

[0004] The technical bottlenecks faced by the above-mentioned four-dimensional variational assimilation system in the regional ocean model cause the utilization rate of computing resources to decline sharply with the expansion of scale, which seriously restricts the operational application of high-resolution ocean numerical forecasting. Summary of the invention

[0005] In view of the deficiencies in the relevant technologies, the purpose of the present invention is to provide a data communication optimization method and system based on a regional ocean model, so as to solve the technical problem that the computing resource utilization rate of the prior art four-dimensional variational assimilation system in the regional ocean model decreases sharply with the expansion of the scale, which seriously restricts the commercial application of high-resolution ocean numerical forecasting.

[0006] The present invention provides a data communication optimization method based on a regional ocean model, comprising the following steps: Data acquisition step: acquiring ocean model variables in a regional ocean model, and defining a data read-in decomposition descriptor and a data read-out decomposition descriptor of the ocean model variables through a parallel input and output system; Mapping relationship configuration step: configuring a data-in mapping relationship between a storage position of the ocean model variable data in a storage file and a storage position in the memory for the data-in decomposition descriptor, and configuring a data-out mapping relationship between a storage position of the ocean model variable data in a storage file and a storage position in the memory for the data-out decomposition descriptor; Data processing steps: reading the calculated data according to the data read-in decomposition descriptor, calculating the calculated data through the four-dimensional variational assimilation system model, and outputting the calculated data according to the data read-out decomposition descriptor.

[0007] The embodiment of the present invention achieves efficient data acquisition and processing of ocean model variables in a regional ocean model by defining data input and output decomposition descriptors, optimizes data communication, reduces data transmission and storage overhead, and improves computing efficiency.

[0008] In some embodiments of the present invention, the mapping relationship configuration step is specifically: The ocean model variable data is divided into multiple data blocks, each data block is assigned to a process for processing, the boundary data of the adjacent processes of the current process are read into the current process through the data read-in decomposition descriptor, and the real data of the current process is directly written into the storage file through the data read-out decomposition descriptor.

[0009] The embodiment of the present invention divides the ocean model variable data into IO data domains, reduces the boundary data exchange between processes, realizes efficient reading and writing of data in parallel computing, further optimizes data communication, and reduces the communication overhead between processes.

[0010] In some embodiments of the present invention, the mapping relationship configuration step further includes: The offset is calculated by looping through the latitude direction through the first layer, and the longitude direction is traversed through the second layer. The data index position is calculated according to the offset, and the mapping relationship is obtained according to the data index position.

[0011] The embodiment of the present invention realizes the calculation of the data index position in the latitude and longitude directions through the inner first layer loop, ensures the accuracy of the data mapping relationship, and improves the precision and efficiency of data processing.

[0012] In some embodiments of the present invention, the calculation model of the offset is:

[0013] in, is the offset; is the latitude direction; is the latitudinal offset of the ocean model variable; is the global domain length in longitude.

[0014] The embodiment of the present invention clarifies the method for calculating the data offset in the latitude direction through the calculation model of the offset, provides a theoretical basis for determining the data index position, and enhances the scientific nature of data processing.

[0015] In some embodiments of the present invention, the calculation model of the data index position is:

[0016] in, is the data index position; is the longitude direction; is the offset of the ocean mode variable in the longitude direction.

[0017] The embodiment of the present invention clarifies the method for calculating the data offset in the longitude direction through the calculation model of the data index position, further improves the configuration of the data mapping relationship, and improves the accuracy of data processing.

[0018] In some embodiments of the present invention, the mapping relationship configuration step further includes: When configuring the data read-in mapping relationship, the range of the latitude direction traversed through the first layer loop is 1 to 1. The range of longitude direction traversed through the second layer loop is 1 to 1; in, , 1. 1. The calculation models of 1 are: ； ； ； ; in, is the number of grid points on the boundary side.

[0019] When calculating the data readout mapping relationship, the range of the latitude direction traversed through the first layer loop is to , the range of longitude direction traversed through the second layer loop is to ; in, , 2. 2. The calculation models of 2 are: ； ； ； .

[0020] The embodiment of the present invention optimizes the configuration of the data read-in domain write-out mapping relationship by limiting the traversal range in the latitude and longitude directions.

[0021] In some embodiments of the present invention, the data processing step is specifically: In the outer loop of the four-dimensional variational assimilation system, the calculation data is loaded and read according to the data read-in decomposition descriptor through the nonlinear model, and the time step calculation is performed according to the calculation data; In the inner loop of the four-dimensional variational assimilation system, the calculation data are loaded and read according to the data reading decomposition descriptor through the adjoint model and the tangent line model, and the time step calculation is performed according to the calculation data.

[0022] The embodiment of the present invention realizes efficient data processing and ensures the accuracy and reliability of the calculation results through time stepping calculation of nonlinear models, adjoint models and tangent linear models.

[0023] In some embodiments of the present invention, the data processing step further includes: The preset configuration file is loaded into the memory, and the calculated data is output through the data read-out decomposition descriptor according to the variable ID and variable generation time configured in the configuration file, and the outer loop and the inner loop are completed according to the total time step and time step configured in the configuration file.

[0024] The embodiment of the present invention realizes automatic data processing by loading configuration files and completing internal and external loops according to the configurations, reduces manual intervention, and improves the efficiency and repeatability of data processing.

[0025] Some embodiments of the present invention further provide a data communication optimization system based on a regional ocean model, comprising: Data acquisition module: acquires ocean model variables in the regional ocean model, and defines the data read-in decomposition descriptor and data read-out decomposition descriptor of the ocean model variables through a parallel input and output system; A mapping relationship configuration module: a data-read-in mapping relationship between a storage position of the ocean model variable data in a storage file and a storage position in a memory is configured for the data-read-in decomposition descriptor, and a data-read-out mapping relationship between a storage position of the ocean model variable data in a storage file and a storage position in a memory is configured for the data-read-out decomposition descriptor; Data processing module: reads the calculated data according to the data read-in decomposition descriptor, calculates the calculated data through the four-dimensional variational assimilation system model, and outputs the calculated data according to the data read-out decomposition descriptor.

[0026] The embodiment of the present invention defines data read-in and read-out decomposition descriptors through a parallel input and output system, thereby realizing efficient acquisition and transmission of ocean model variables, reducing data communication overhead, and utilizing a four-dimensional variational assimilation system model to process the read calculation data, thereby being able to efficiently complete complex ocean model calculation tasks, supporting multi-process parallel processing, being able to effectively utilize computing resources, and reducing program running time. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work. Figure 1 A flow chart of a data communication optimization method based on a regional ocean model provided by an embodiment of the present invention; Figure 2 A block diagram of data reading provided by an embodiment of the present invention; Figure 3 A block diagram of data reading provided by an embodiment of the present invention; Figure 4 A schematic diagram of the offset of an ocean model variable in the i direction and the j direction provided in an embodiment of the present invention; Figure 5 An operation flow chart of a data communication optimization method based on a regional ocean model provided by an embodiment of the present invention; Figure 6 A schematic structural diagram of a data communication optimization system based on a regional ocean model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application. It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. Four-dimensional variational assimilation for regional ocean models (4D-Var) is a data assimilation method used in numerical weather prediction and ocean models. The goal of 4D-Var is to generate an optimal initial condition to improve the accuracy of the forecast by combining the observed data and the numerical model. The 4D-Var system works by minimizing a cost function that measures the difference between the model forecast and the observed data.

[0029] The steps of the four-dimensional variational assimilation system include: Forward Integration: Starting from initial conditions, run the numerical model forward and integrate to generate model forecasts.

[0030] Adjoint mode: The gradient of the cost function with respect to the initial condition is calculated by reverse integration of the adjoint mode.

[0031] Tangent linear mode: used to approximate the behavior of nonlinear models and help calculate gradients.

[0032] Optimization process: Through the iterative optimization process, the initial conditions are adjusted to minimize the cost function.

[0033] In regional ocean models, 4D variational assimilation systems are often used for ocean numerical simulations, especially in high-resolution and large-scale computing scenarios. However, a major challenge of 4D variational assimilation systems is their high computational complexity, especially in a multi-process parallel architecture, where communication overhead becomes a bottleneck.

[0034] The PIO (Parallel Input / Output) system is a parallel input / output system for high-performance computing, designed to optimize file read and write operations in large-scale parallel computing. In the regional ocean model, the PIO system is used to manage the read and write operations of model data, especially in a multi-process parallel computing environment. The PIO system allows multiple processes to read and write data in parallel, thereby reducing the waiting time of I / O operations.

[0035] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other. The technical solution of the present invention is described in detail below in conjunction with specific embodiments and the accompanying drawings.

[0036] As attached Figure 1 As shown, the present invention provides a data communication optimization method based on regional ocean model, comprising the following steps: Data acquisition step S1: acquiring ocean model variables in a regional ocean model, defining a data read-in decomposition descriptor and a data read-out decomposition descriptor of the ocean model variables through a parallel input and output system; optionally, the ocean model variable types include psi, rho, u, and v in 2D / 3D; Where psi represents the stream function in the ocean, which is used to describe the horizontal flow in the ocean. The stream function is a scalar field whose gradient can be used to represent the velocity field. In the regional ocean model, psi is used to calculate the horizontal velocity (u and v).

[0037] Rho represents the density of seawater and is a 3D variable that represents the density distribution of seawater at different depths and locations. Density is one of the key variables in ocean models and directly affects the vertical movement and stratification stability in the ocean. Changes in density are usually determined by factors such as temperature, salinity, and pressure.

[0038] u represents the eastward flow velocity in the ocean, that is, the velocity component of seawater in the east-west direction. It is a 3D variable that represents the eastward flow velocity at different depths and locations. u is one of the important variables describing the horizontal flow of the ocean and is usually used together with v (northward flow velocity) to calculate the horizontal circulation in the ocean.

[0039] v represents the northward flow velocity in the ocean, that is, the velocity component of seawater in the north-south direction. It is a 3D variable that represents the northward flow velocity at different depths and locations. v is used together with u to describe horizontal flow in the ocean and calculate ocean circulation and momentum equations.

[0040] Optionally, the data read-in decomposition descriptor and the data read-out decomposition descriptor both define the storage location of the ocean model variable data in the file, the storage location of the ocean model variable data in the memory, and the block method of the ocean model variable data.

[0041] Mapping relationship configuration step S2: configuring a data read-in mapping relationship between a storage location of the ocean model variable data in a storage file and a storage location in the memory for the data read-in decomposition descriptor, and configuring a data read-out mapping relationship between a storage location of the ocean model variable data in a storage file and a storage location in the memory for the data read-out decomposition descriptor; Data processing step S3: Read the calculated data according to the data read-in decomposition descriptor, calculate the calculated data through the four-dimensional variational assimilation system model, and output the calculated data according to the data read-out decomposition descriptor.

[0042] Based on the above method, by defining data read-in and read-out decomposition descriptors, efficient data acquisition and processing of ocean model variables in regional ocean models are achieved, data communication is optimized, the overhead of data transmission and storage is reduced, and the computational efficiency is improved.

[0043] Combination Figure 2-Figure 3 As shown, the mapping relationship configuration step S2 is specifically as follows: The ocean model variable data is divided into multiple data blocks, each data block is assigned to a process for processing, the boundary data of the adjacent processes of the current process is read into the data block of the current process through the data read-in decomposition descriptor, and the data executed by the current process is written into the storage file, and the data executed by the current process is directly written into the storage file through the data read-out decomposition descriptor.

[0044] Among them, the above-mentioned blocking strategy of reading different data into data blocks and writing them into storage files according to the data read-in blocking descriptor and the data read-out blocking descriptor respectively is a dual-mode PIO data blocking strategy designed by the present invention, and the dual-mode PIO data blocking strategy includes a data read-out strategy and a data read-in strategy.

[0045] like Figure 2 As shown, the data reading strategy is to only absorb the real data of the current process into the storage file according to business needs.

[0046] like Figure 3 As shown in the figure, the data reading strategy is that the boundary data of the current process is needed after data preprocessing. Therefore, when the data is divided into blocks, the boundary data of the adjacent processes of the current process are read into the data block of the current process. If there is no adjacent process, there is no need to include the boundary data on this side, such as the process data blocks at the lower right boundary and the upper left boundary. Figure 3 As shown in the lower left corner. Through the data reading strategy, the adjacent process boundary data is read into the data block of the current process in advance when reading data, and the inter-process communication steps for boundary data exchange after data preprocessing are optimized, thereby reducing a large number of inter-process communication operations.

[0047] In some embodiments of the present invention, the mapping relationship configuration step S2 further includes: The offset is calculated by looping through the latitude direction through the first layer, and the longitude direction is traversed through the second layer. The data index position is calculated according to the offset, and the mapping relationship is obtained according to the data index position.

[0048] The calculation model of the offset is:

[0049] in, is the offset; is the latitude direction; is the latitudinal offset of the ocean model variable; is the global domain length in longitude.

[0050] in, is the offset of the starting position of the ocean mode variable when traversing the j direction, There is no offset in the i direction superimposed.

[0051] The calculation model of data index position is:

[0052] in, is the data index position; is the longitude direction; is the offset of the ocean mode variable in the longitude direction.

[0053] in, The i-direction offset is superimposed on the j-direction offset, that is, the final offset of each point in the ocean mode variable. The mapping relationship between the storage location in the file and the storage location in the memory is stored in a one-dimensional array. The index of the array is the storage location in the memory, and the value in the index is the storage location in the file. The mapping relationship is established through the one-dimensional array.

[0054] Alternatively, if Figure 4 As shown, the offset of rho type variables and u type variables in the j direction is 0, and the offset of v type and psi type variables in the j direction is 1.

[0055] The offset of rho type variables and v type variables in the i direction is 0, and the offset of psi type variables and u type variables in the i direction is 1.

[0056] Through the calculation model of offset and data index position, the calculation method of data offset in latitude and longitude directions is clarified, the configuration of data mapping relationship is further improved, and the accuracy of data processing is improved.

[0057] In some embodiments of the present invention, the mapping relationship configuration step S2 further includes: When configuring data to read the mapping relationship, the range of latitude direction traversed through the first layer loop is 1 to 1. The range of longitude direction traversed through the second layer loop is 1 to 1; in, , 1. 1. The calculation models of 1 are: ； ； ； ; in, is the number of grid points on the boundary side.

[0058] When calculating the data readout mapping relationship, the range of the latitude direction traversed through the first layer loop is to , the range of longitude direction traversed through the second layer loop is to ; in, , 2. 2. The calculation models of 2 are: ； ； ； .

[0059] in, , 2. 2. 2Excludes data in border areas.

[0060] By limiting the traversal range in latitude and longitude directions, the configuration of data reading mapping relationship is optimized, unnecessary computing overhead is reduced, and the efficiency of data processing is improved.

[0061] Combination Figure 5 As shown, the data processing step S3 is specifically as follows: In the outer loop of the four-dimensional variational assimilation system, the calculation data is loaded and read according to the data read-in decomposition descriptor through the nonlinear model, the calculation data is preprocessed, and the preprocessed calculation data is time-stepped; optionally, the preprocessing includes time interpolation.

[0062] In the inner loop of the four-dimensional variational assimilation system, the calculation data are loaded and read according to the data reading decomposition descriptor through the adjoint model and the tangent linear model, and time step calculation is performed according to the calculation data.

[0063] The preset configuration file is loaded into the memory, and the calculated data is output through the data read-out decomposition descriptor according to the variable ID and variable generation time configured in the configuration file, and the outer loop and the inner loop are completed according to the total time step and time step configured in the configuration file.

[0064] Optionally, the configuration file is obtained according to business needs. The configuration file .in is loaded into the memory. The configuration file is configured with parameters related to the running time, such as NTIMES (total number of time steps), DT (time step), Nouter (number of outer loops), Ninner (number of inner loops); Hout (variable id) controls the variables output to the history file, and NHIS is used to control the time step of generating the history file.

[0065] The outer loop and inner loop are completed according to the total time step and time step configured in the configuration file. The total number of steps of the nonlinear model, adjoint model and tangent model is obtained through NTIMES in the configuration file, DT is the iteration frequency, the total number of steps is the number of iterations, and DT*NTIMES is the total simulation time. If DT is 45 seconds and NTIMES is 1920, the nonlinear model will iterate once every 45 seconds, for a total of 1920 iterations. After the nonlinear model is completed, the inner loop will be entered. The inner loop will first calculate the adjoint model. After the adjoint model calculation reaches the total number of steps, the tangent model calculation will be performed. After the tangent model calculation reaches the total number of steps, it will be determined whether the inner loop number has been reached. If not, the calculation of the adjoint model and the calculation of the tangent model will continue. If the inner loop number has been reached, check whether the outer loop has been completed. If not, the calculation of the nonlinear model will continue until both the inner and outer loops are completed.

[0066] Through the time-stepping calculation of nonlinear models, adjoint models and tangent models, efficient data processing is achieved, ensuring the accuracy and reliability of the calculation results. By loading the configuration file and completing the internal and external loops according to the configuration, automatic data processing is achieved, reducing manual intervention and improving the efficiency and repeatability of data processing.

[0067] like Figure 6 As shown, an embodiment of the present invention further provides a data communication optimization system based on a regional ocean model, comprising: Data acquisition module 1: obtains ocean model variables in the regional ocean model, and defines the data read-in decomposition descriptor and data read-out decomposition descriptor of the ocean model variables through a parallel input and output system; Mapping relationship configuration module 2: configures the data read-in mapping relationship between the storage location of the ocean model variable data in the storage file and the storage location in the memory for the data read-in decomposition descriptor, and configures the data read-out mapping relationship between the storage location of the ocean model variable data in the storage file and the storage location in the memory for the data read-out decomposition descriptor; Data processing module 3: reads the calculated data according to the data read-in decomposition descriptor, calculates the calculated data through the four-dimensional variational assimilation system model, and outputs the calculated data according to the data read-out decomposition descriptor.

[0068] By defining data read-in and read-out decomposition descriptors through a parallel input and output system, efficient acquisition and transmission of ocean model variables are achieved, reducing data communication overhead. The four-dimensional variational assimilation system model is used to process the read calculation data, which can efficiently complete complex ocean model calculation tasks, support multi-process parallel processing, effectively utilize computing resources, and reduce program running time.

[0069] It should be noted that the above is a reference for a data communication optimization method and system based on a regional ocean model, and the present invention is not limited thereto.

[0070] The embodiments of the present invention realize efficient data acquisition and processing of ocean model variables in a regional ocean model by defining data read-in and read-out decomposition descriptors, optimize data communication, reduce data transmission, and improve computing efficiency, thereby solving the technical problem that the computing resource utilization rate of the prior art four-dimensional variational assimilation system in a regional ocean model decreases sharply with the expansion of scale, which seriously restricts the commercial application of high-resolution ocean numerical forecasting.

[0071] Finally, it should be noted that: the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other. The above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention, which should be included in the scope of the technical solution for protection of the present invention.

Claims

1. A data communication optimization method based on regional ocean model, characterized in that: The steps include: Data acquisition step: acquiring ocean model variables in a regional ocean model, and defining a data read-in decomposition descriptor and a data read-out decomposition descriptor of the ocean model variables through a parallel input and output system; Mapping relationship configuration step: configuring a data-in mapping relationship between a storage position of the ocean model variable data in a storage file and a storage position in the memory for the data-in decomposition descriptor, and configuring a data-out mapping relationship between a storage position of the ocean model variable data in a storage file and a storage position in the memory for the data-out decomposition descriptor; Data processing steps: reading the calculated data according to the data read-in decomposition descriptor, calculating the calculated data through the four-dimensional variational assimilation system model, and outputting the calculated data according to the data read-out decomposition descriptor.

2. The data communication optimization method based on regional ocean model according to claim 1, characterized in that: The mapping relationship configuration steps are specifically as follows: The ocean model variable data is divided into multiple data blocks, each data block is assigned to a process for processing, the boundary data of the adjacent processes of the current process are read into the current process through the data read-in decomposition descriptor, and the real data of the current process is directly written into the storage file through the data read-out decomposition descriptor.

3. The data communication optimization method based on regional ocean model according to claim 2 is characterized in that: The mapping relationship configuration step also includes: The offset is calculated by looping through the latitude direction through the first layer, and the longitude direction is traversed through the second layer, and the data index position is calculated according to the offset, and the mapping relationship is obtained according to the data index position.

4. The data communication optimization method based on regional ocean model according to claim 3 is characterized in that: The calculation model of the offset is: in, is the offset; is the latitude direction; is the latitudinal offset of the ocean model variable; is the global domain length in longitude.

5. The data communication optimization method based on regional ocean model according to claim 4 is characterized in that: The calculation model of the data index position is: in, is the data index position; is the longitude direction; is the offset of the ocean mode variable in the longitude direction.

6. The data communication optimization method based on regional ocean model according to claim 4 is characterized in that: The mapping relationship configuration step also includes: When configuring the data read-in mapping relationship, the range of the latitude direction traversed through the first layer loop is 1 to 1. The range of longitude direction traversed through the second layer loop is 1 to 1; in, , 1.

1. The calculation models of 1 are: ； ； ； ； in, is the number of grid points on the boundary side.

7. The data communication optimization method based on regional ocean model according to claim 5, characterized in that: The mapping relationship configuration step also includes: When calculating the data readout mapping relationship, the range of the latitude direction traversed through the first layer loop is to , the range of longitude direction traversed through the second layer loop is to ; in, , 2.

2. The calculation models of 2 are: ； ； ； 。 8. The data communication optimization method based on regional ocean model according to claim 7, characterized in that: The data processing steps are specifically as follows: In the outer loop of the four-dimensional variational assimilation system, the calculation data is loaded and read according to the data read-in decomposition descriptor through the nonlinear model, and the time step calculation is performed according to the calculation data; In the inner loop of the four-dimensional variational assimilation system, the calculation data are loaded and read according to the data reading decomposition descriptor through the adjoint model and the tangent line model, and the time step calculation is performed according to the calculation data.

9. The data communication optimization method based on regional ocean model according to claim 8, characterized in that: The data processing step also includes: The preset configuration file is loaded into the memory, and the calculated data is output through the data read-out decomposition descriptor according to the variable ID and variable generation time configured in the configuration file, and the outer loop and the inner loop are completed according to the total time step and time step configured in the configuration file.

10. A data communication optimization system based on regional ocean model, characterized in that: include: Data acquisition module: acquires ocean model variables in the regional ocean model, and defines the data read-in decomposition descriptor and the data read-out decomposition descriptor of the ocean model variables through a parallel input and output system; A mapping relationship configuration module: a data-read-in mapping relationship between a storage position of the ocean model variable data in a storage file and a storage position in a memory is configured for the data-read-in decomposition descriptor, and a data-read-out mapping relationship between a storage position of the ocean model variable data in a storage file and a storage position in a memory is configured for the data-read-out decomposition descriptor; Data processing module: reads the calculated data according to the data read-in decomposition descriptor, calculates the calculated data through the four-dimensional variational assimilation system model, and outputs the calculated data according to the data read-out decomposition descriptor.

Images