A data communication optimization method and system based on a regional ocean model

Abstract

The present invention relates to a data communication optimization method and system based on a regional ocean model, belonging to the technical field of data communication, including obtaining ocean model variables in the regional ocean model, defining a read-in decomposition descriptor and a read-out decomposition descriptor for the data of the ocean model variables through a parallel input / output system; respectively configuring a data read-in mapping relationship and a 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-in decomposition descriptor and the data read-out decomposition descriptor; reading and calculating data according to the data read-in decomposition descriptor, calculating the calculated data through a four-dimensional variational assimilation system model, and outputting the calculated data according to the data read-out decomposition descriptor. By defining the data read-in and read-out decomposition descriptors, efficient data acquisition and processing of ocean model variables in the regional ocean model are realized, data communication is optimized, the overhead of data transmission between processes is reduced, and the calculation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data communication, and particularly 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 applied to ocean numerical simulation, capable of simulating the physical, chemical, and biological processes of the ocean. To improve the simulation accuracy and prediction ability, an 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 ultra-large-scale computing scenarios, the four-dimensional variational assimilation system still faces some technical bottlenecks, such as:

[0003] The regional ocean model adopts a horizontal domain decomposition parallel strategy, and each sub-domain needs to exchange boundary data with eight neighboring processes through the Message Passing Interface. In the four-dimensional variational assimilation system, the nested loops (outer loop and inner loop) of the forward integration, adjoint model, and tangent linear process lead to an O(N²)-order increase in the number of communications, seriously affecting the computing efficiency.

[0004] With the expansion of the process scale, the number of boundaries of the two-dimensional domain decomposition increases with the number of processes. The point-to-point communication links compete for the physical bandwidth, and the network congestion and delay amplification effects are significant. In a high-performance computing cluster environment, the traditional communication mode is difficult to maintain linear scalability, restricting the application of the four-dimensional variational assimilation system in large-scale computing.

[0005] The above technical bottlenecks faced by the four-dimensional variational assimilation system in the regional ocean model lead to a sharp decline in the utilization rate of computing resources with the increase in scale, severely restricting the operational application of high-resolution ocean numerical forecasting. Summary of the Invention

[0006] Aiming at the deficiencies in the related technologies, the purpose of the present invention is to provide a data communication optimization method and system based on a regional ocean model to solve the technical problem that in the existing technology, in the regional ocean model of the four-dimensional variational assimilation system, the utilization rate of computing resources sharply decays with the increase in scale, severely restricting the operational application of high-resolution ocean numerical forecasting.

[0007] The present invention provides a data communication optimization method based on a regional ocean model, including the following steps:

[0008] Data acquisition step: Acquire ocean model variables in the regional ocean model, and define the read-in decomposition descriptor and data read-out decomposition descriptor of the data of the ocean model variables through a parallel input / output system;

[0009] Mapping relationship configuration steps: Configure the data reading mapping relationship between the storage location of ocean model variable data in the storage file and the storage location in memory for the data reading decomposition descriptor, and configure the data reading mapping relationship between the storage location of ocean model variable data in the storage file and the storage location in memory for the data reading decomposition descriptor;

[0010] Data processing steps: Read the calculation data according to the data reading decomposition descriptor, perform calculations on the calculation data through the four-dimensional variational assimilation system model, and output the calculated data according to the data reading decomposition descriptor.

[0011] In the embodiments of the present invention, by defining the data reading and reading decomposition descriptors, efficient data acquisition and processing of ocean model variables in the regional ocean model are realized, data communication is optimized, the overhead of data transmission and storage is reduced, and the calculation efficiency is improved.

[0012] In some embodiments of the present invention, the mapping relationship configuration steps are specifically as follows:

[0013] Divide the ocean model variable data into multiple data blocks, allocate a process to process each data block, read the boundary data of the adjacent process of the current process into the current process through the data reading decomposition descriptor, and directly write the real data of the current process into the storage file through the data reading decomposition descriptor.

[0014] In the embodiments of the present invention, by dividing the ocean model variable data into IO data domains, the exchange of boundary data between processes is reduced, efficient data reading and writing in parallel computing are realized, data communication is further optimized, and the communication overhead between processes is reduced.

[0015] In some embodiments of the present invention, the mapping relationship configuration steps further include:

[0016] Calculate the offset by traversing the latitude direction through the first-layer loop, traverse the longitude direction through the second-layer loop, calculate the data index position according to the offset, and obtain the mapping relationship according to the data index position.

[0017] In the embodiments of the present invention, through the first-layer loop, the calculation of the data index position in the latitude and longitude directions is realized, the accuracy of the data mapping relationship is ensured, and the accuracy and efficiency of data processing are improved.

[0018] In some embodiments of the present invention, the calculation model of the offset is:

[0019]

[0020] Where is the offset; is the latitude direction; is the offset of the ocean model variable in the latitude direction; is the global domain length in the longitude direction.

[0021] In the embodiments of the present invention, through the calculation model of the offset, the calculation method of the data offset in the latitude direction is clarified, providing a theoretical basis for the determination of the data indexing position and enhancing the scientific nature of data processing.

[0022] In some embodiments of the present invention, the calculation model of the data indexing position is:

[0023]

[0024] where is the data indexing position; is the longitude direction; is the offset of the ocean model variable in the longitude direction.

[0025] In the embodiments of the present invention, through the calculation model of the data indexing position, the calculation method of the data offset in the longitude direction is clarified, further improving the configuration of the data mapping relationship and enhancing the accuracy of data processing.

[0026] In some embodiments of the present invention, the mapping relationship configuration step further includes:

[0027] When configuring the data read-in mapping relationship, the range of the latitude direction is traversed through the first-layer loop as from 1 to 1, and the range of the longitude direction is traversed through the second-layer loop as from 1 to 1;

[0028] where , 1, 1, 1 of the calculation models are respectively:

[0029] ；

[0030] ；

[0031] ；

[0032] ;

[0033] where is the number of grid points on the boundary side.

[0034] When calculating the data read-out mapping relationship, the range of the latitude direction is traversed through the first-layer loop as to through the second - layer loop, the traversal range in the longitude direction is from to ;

[0035] Among them, and 2, 2, the calculation models of 2 are respectively:

[0036] ；

[0037] ；

[0038] ；

[0039] .

[0040] By limiting the traversal ranges in the latitude and longitude directions in the embodiments of the present invention, the configuration of the mapping relationship between data reading and domain writing is optimized.

[0041] In some embodiments of the present invention, the data - processing steps are specifically as follows:

[0042] In the outer loop of the four - dimensional variational assimilation system, the calculation data is loaded and read according to the data - reading decomposition descriptor through a non - linear model, and time - stepping calculation is performed based on the calculation data;

[0043] In the inner loop of the four - dimensional variational assimilation system, the calculation data is loaded and read according to the data - reading decomposition descriptor through the adjoint model and the tangent - linear model respectively, and time - stepping calculation is performed based on the calculation data.

[0044] By the time - stepping calculations of the non - linear model, the adjoint model and the tangent - linear model in the embodiments of the present invention, efficient data processing is achieved, ensuring the accuracy and reliability of the calculation results.

[0045] In some embodiments of the present invention, the data - processing steps further include:

[0046] A preset configuration file is loaded into the memory, the calculated data is output according to the variable id and the variable generation time configured in the configuration file through the data - reading decomposition descriptor, and the outer loop and the inner loop are completed according to the total time steps and the time step length configured in the configuration file.

[0047] By loading the configuration file and completing the inner and outer loops according to the configuration in the embodiments of the present invention, automated data processing is achieved, reducing manual intervention and improving the efficiency and repeatability of data processing.

[0048] Some embodiments of the present invention further provide a data communication optimization system based on a regional ocean model, including:

[0049] A data acquisition module: acquires ocean model variables in the regional ocean model, and defines a read-in decomposition descriptor and a read-out decomposition descriptor for the data of the ocean model variables through a parallel input / output system;

[0050] A mapping relationship configuration module: configures a 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 a 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;

[0051] A data processing module: reads calculation data according to the data read-in decomposition descriptor, performs calculations on the calculation data through a four-dimensional variational assimilation system model, and outputs the calculated data according to the data read-out decomposition descriptor.

[0052] In the embodiments of the present invention, a read-in and read-out decomposition descriptor for data is defined through a parallel input / output system, realizing the efficient acquisition and transmission of ocean model variables, reducing the overhead of data communication, processing the read calculation data by using a four-dimensional variational assimilation system model, being able to efficiently complete complex ocean model calculation tasks, supporting multi-process parallel processing, being able to effectively utilize computing resources, and reducing the program running time. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will describe in detail the specific embodiments of the present invention with reference to the drawings. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0054] Figure 1 It is a flowchart of a data communication optimization method based on a regional ocean model provided by an embodiment of the present invention;

[0055] Figure 2 It is a schematic diagram of a block for data read-out provided by an embodiment of the present invention;

[0056] Figure 3 It is a schematic diagram of a block for data read-in provided by an embodiment of the present invention;

[0057] Figure 4 It is a schematic diagram of the offsets of an ocean model variable in the i direction and the j direction provided by an embodiment of the present invention;

[0058] Figure 5A flowchart of the operation of a data communication optimization method based on a regional ocean model provided by an embodiment of the present invention;

[0059] Figure 6 A schematic structural diagram of a data communication optimization system based on a regional ocean model provided by an embodiment of the present invention. Detailed implementation manners

[0060] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be described and explained below with reference to 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 used to limit the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments provided in the present application without making creative efforts belong to the scope of protection of the present application.

[0061] It should be noted that the terms used herein are only for describing the specific implementation manners 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 "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0062] Four-dimensional variational assimilation (4D-Var) for regional ocean models is a data assimilation method for numerical weather forecasting and ocean models. The goal of four-dimensional variational assimilation is to generate an optimal initial condition by combining observational data and numerical models to improve the accuracy of forecasts. The four-dimensional variational assimilation system works by minimizing a cost function that measures the difference between the model forecast and the observational data.

[0063] Among them, the steps of the four-dimensional variational assimilation system include:

[0064] Forward integration: Starting from the initial condition, run the numerical model for forward integration to generate a model forecast.

[0065] Adjoint model: Calculate the gradient of the cost function with respect to the initial condition by integrating backward through the adjoint model.

[0066] Tangent linear model: Used to approximate the behavior of the nonlinear model and help calculate the gradient.

[0067] Optimization process: Adjust the initial condition through an iterative optimization process to minimize the cost function.

[0068] In the regional ocean model, the four-dimensional variational assimilation system is usually used for ocean numerical simulation, especially in high-resolution and large-scale computing scenarios. However, a major challenge of the four-dimensional variational assimilation system is its high computational complexity. Especially in a multi-process parallel architecture, the communication overhead becomes a bottleneck.

[0069] The PIO (Parallel Input / Output) system is a parallel input-output system for high-performance computing, aiming 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, thus reducing the waiting time for I / O operations.

[0070] Without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0071] The following combines specific embodiments and the accompanying drawings of the specification to elaborate on the technical solutions of the present invention in detail.

[0072] As shown in the Figure 1 accompanying drawings, the present invention provides a data communication optimization method based on a regional ocean model, including the following steps:

[0073] Data acquisition step S1: Acquire ocean model variables in the regional ocean model, and define the read-in decomposition descriptor and data read-out decomposition descriptor of the data of the ocean model variables through the parallel input-output system; optionally, the types of ocean model variables include psi, rho, u, v in 2D / 3D;

[0074] Among them, 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, and its gradient can be used to represent the velocity field. In the regional ocean model, psi is used to calculate the horizontal flow velocity (u and v).

[0075] rho represents the density of seawater, which is a 3D variable representing the density distribution of seawater at different depths and positions. Density is one of the key variables in the ocean model, directly affecting the vertical movement and stratification stability in the ocean. The change of density is usually determined by factors such as temperature, salinity, and pressure.

[0076] u represents the eastward flow velocity in the ocean, that is, the velocity component of seawater in the east-west direction, which is a 3D variable representing the eastward flow velocity at different depths and positions. u is one of the important variables describing the horizontal flow in the ocean, and is usually used together with v (northward flow velocity) to calculate the horizontal circulation in the ocean.

[0077] v represents the northward velocity in the ocean, that is, the velocity component of seawater in the north-south direction. It is a 3D variable representing the northward velocity at different depths and positions. Together with u, v is used to describe the horizontal flow in the ocean and calculate the ocean circulation and momentum equations.

[0078] Optionally, both the data read-in decomposition descriptor and the data read-out decomposition descriptor define the storage location of ocean model variable data in the file, the storage location of ocean model variable data in memory, and the way of chunking the ocean model variable data.

[0079] Mapping relationship configuration step S2: Configure the data read-in mapping relationship between the storage location of ocean model variable data in the storage file and the storage location in memory for the data read-in decomposition descriptor, and configure the data read-out mapping relationship between the storage location of ocean model variable data in the storage file and the storage location in memory for the data read-out decomposition descriptor;

[0080] Data processing step S3: Read the calculation data according to the data read-in decomposition descriptor, calculate the calculation data through the four-dimensional variational assimilation system model, and output the calculated data according to the data read-out decomposition descriptor.

[0081] Based on the above method, by defining the data read-in and read-out decomposition descriptors, efficient data acquisition and processing of ocean model variables in the regional ocean model are realized, data communication is optimized, the overhead of data transmission and storage is reduced, and the calculation efficiency is improved.

[0082] Combined Figures 2 - 3 As shown in

[0083] Divide the ocean model variable data into multiple data blocks, allocate a process to handle each data block, read the boundary data of the adjacent processes of the current process into the data block of the current process through the data read-in decomposition descriptor, and write the data executed by the current process into the storage file. Write the data executed by the current process directly into the storage file through the data read-out decomposition descriptor.

[0084] Among them, the above chunking strategies of reading different data into data blocks and writing them into the storage file according to the data read-in decomposition descriptor and the data read-out decomposition descriptor respectively are the dual-mode PIO data chunking strategies designed by the present invention. The dual-mode PIO data chunking strategies include a data read-out strategy and a data read-in strategy.

[0085] As Figure 2 shown, the data read-out strategy is to suck only the real data of the current process into the storage file according to business requirements.

[0086] As Figure 3As shown, the data reading strategy is that since the boundary data of this process is required after data preprocessing, when dividing data into blocks, the boundary data of adjacent processes of the current process is read into the data block of the current process. For processes without adjacent processes, the boundary data on this side does not need to be included, such as the process data blocks at the lower right boundary and the upper left boundary. The boundary data is as Figure 3 shown in the lower left corner. Through the data reading strategy, when reading data, the boundary data of adjacent processes is read into the data block of the current process in advance, optimizing the inter-process communication step for boundary data exchange after data preprocessing, thereby reducing a large number of inter-process communication operations.

[0087] In some embodiments of the present invention, the mapping relationship configuration step S2 further includes:

[0088] Calculating the offset through the first-layer loop traversing the latitude direction, traversing the longitude direction through the second-layer loop, calculating the data index position according to the offset, and obtaining the mapping relationship according to the data index position.

[0089] Among them, the calculation model of the offset is:

[0090]

[0091] Among them, is the offset; is the latitude direction; is the offset of the ocean model variable in the latitude direction; is the global domain length in the longitude direction.

[0092] Among them, is the offset of the starting position of the ocean model variable when traversing the j direction, without superimposing the offset in the i direction.

[0093] The calculation model of the data index position is:

[0094]

[0095] Among them, is the data index position; is the longitude direction; is the offset of the ocean model variable in the longitude direction.

[0096] Among them, is the offset in the i direction superimposed on the basis of the offset in the j direction, that is, the final offset of each point in the ocean model variable. The mapping relationship between the storage position in the file and the storage position in the memory is stored in a one-dimensional array. The index of the array is the storage position in the memory, and the value in the index is the storage position in the file. The mapping relationship is established through the one-dimensional array.

[0097] Optionally, asFigure 4 As shown, the offset of the rho - type variable and the u - type variable in the j - direction is 0, and the offset of the v - type and psi - type variables in the j - direction is 1.

[0098] The offset of the rho - type variable and the v - type variable in the i - direction is 0, and the offset of the psi - type variable and the u - type variable in the i - direction is 1.

[0099] Through the calculation model of the offset and the data index position, the calculation methods of the data offsets in the latitude direction and the longitude direction are clarified, further improving the configuration of the data mapping relationship and enhancing the accuracy of data processing.

[0100] In some embodiments of the present invention, the mapping relationship configuration step S2 further includes:

[0101] When configuring the data read - in mapping relationship, through the first - layer loop, the range traversed in the latitude direction is 1 to 1, and through the second - layer loop, the range traversed in the longitude direction is 1 to 1;

[0102] Among them, , 1, 1, 1 of the calculation models are respectively:

[0103] ；

[0104] ；

[0105] ；

[0106] ;

[0107] Among them, is the number of grid points on the boundary side.

[0108] When calculating the data read - out mapping relationship, through the first - layer loop, the range traversed in the latitude direction is to , and through the second - layer loop, the range traversed in the longitude direction is to ;

[0109] Among them, , 2, 2, 2 of the calculation models are respectively:

[0110] ；

[0111] ；

[0112] ；

[0113] 。

[0114] Among them, 、 2, 2, 2 does not include the data in the boundary region.

[0115] By limiting the traversal range in the latitude and longitude directions, the configuration of the data read-in mapping relationship is optimized, unnecessary computational overhead is reduced, and the efficiency of data processing is improved.

[0116] Combined with Figure 5 as shown, the data processing step S3 is specifically as follows:

[0117] In the outer loop of the four-dimensional variational assimilation system, according to the non-linear model, the computational data is loaded and read according to the data read-in decomposition descriptor, the computational data is preprocessed, and the time-stepping calculation is performed on the preprocessed computational data; optionally, the preprocessing includes time interpolation.

[0118] In the inner loop of the four-dimensional variational assimilation system, according to the adjoint model and the tangent linear model respectively, the computational data is loaded and read according to the data read-in decomposition descriptor, and the time-stepping calculation is performed according to the computational data.

[0119] Load the preset configuration file into the memory, and output the calculated data according to the variable id and the variable generation time configured in the configuration file through the data read-out decomposition descriptor, and complete the outer loop and the inner loop according to the total number of time steps and the time step length configured in the configuration file.

[0120] Optionally, the configuration file is obtained according to the service requirements. Load the configuration file.in file 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 length), Nouter (number of outer loops), Ninner (number of inner loops), etc.; Hout (variable id) controls the variables output to the history file, and NHIS is used to control the time step for generating the history file.

[0121] The outer loop and the inner loop are completed according to the total number of time steps and the time step size configured in the configuration file. Among them, the total number of steps of the non - linear model, the adjoint model, and the tangent linear 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. For example, if DT is 45 seconds and NTIMES is 1920, the non - linear model iterates once every 45s for a total of 1920 iterations. After the non - linear model ends, it enters the inner loop. In the inner loop, the adjoint model is calculated first. After the adjoint model calculation reaches the total number of steps, the tangent linear model is calculated. After the tangent linear model calculation reaches the total number of steps, it is judged whether the inner loop count is reached. If the inner loop count is not reached, the calculation of the adjoint model and the tangent linear model continues; if the inner loop count is reached, it is checked whether the outer loop is completed. If the outer loop count is not reached, the calculation of the non - linear model continues until both the inner loop and the outer loop are calculated and completed.

[0122] Through the time - stepping calculations of the non - linear model, the adjoint model, and the tangent linear model, efficient data processing is achieved, ensuring the accuracy and reliability of the calculation results. By loading the configuration file and completing the inner and outer loops according to the configuration, automated data processing is realized, reducing manual intervention and improving the efficiency and repeatability of data processing.

[0123] As Figure 6 shown, the embodiment of the present invention also provides a data communication optimization system based on a regional ocean model, including:

[0124] Data acquisition module 1: Acquire ocean model variables in the regional ocean model, and define the read - in decomposition descriptor and the read - out decomposition descriptor of the data of the ocean model variables through a parallel input - output system.

[0125] Mapping relationship configuration module 2: Configure 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 configure 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.

[0126] Data processing module 3: Read the calculation data according to the data read - in decomposition descriptor, calculate the calculation data through a four - dimensional variational assimilation system model, and output the calculated data according to the data read - out decomposition descriptor.

[0127] By defining the data read - in and read - out decomposition descriptors through a parallel input - output system, the efficient acquisition and transmission of ocean model variables are realized, reducing the overhead of data communication. Using the four - dimensional variational assimilation system model to process the read calculation data can efficiently complete complex ocean model calculation tasks, support multi - process parallel processing, effectively utilize computing resources, and reduce the program running time.

[0128] It should be noted that the above is a reference method for optimizing data communication of a regional ocean model-based data communication optimization method and system, and the present invention is not limited thereto.

[0129] The embodiments of the present invention achieve 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, improve calculation efficiency, and solve the technical problem that in the existing four-dimensional variational assimilation system in a regional ocean model, the utilization rate of computing resources decays sharply with the increase in scale, seriously restricting the operational application of high-resolution ocean numerical forecasting.

[0130] Finally, it should be noted that the various embodiments in this specification are described in a progressive manner, and the key points of each embodiment are the differences from other embodiments. For the same and similar parts between the various embodiments, reference can be made to each other.

[0131] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that: still modifications can be made to the specific implementation manners of the present invention or equivalent replacements can be made to some technical features; without departing from the spirit of the technical solutions of the present invention, they should all be covered within the scope of the technical solutions claimed by the present invention.

Claims

1. An optimization method for data communication based on a regional ocean model, characterized in that It includes the following steps: Data acquisition step: Acquire ocean model variables in the regional ocean model, and define the read-in decomposition descriptor and the read-out decomposition descriptor of the data of the ocean model variables through the parallel input / output system; wherein, the read-in decomposition descriptor and the read-out decomposition descriptor define the storage location of the data of the ocean model variables in the file, the storage location of the data of the ocean model variables in the memory, and the chunking method of the data of the ocean model variables; Mapping relationship configuration step: Configure 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 read-in decomposition descriptor, and configure 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 read-out decomposition descriptor; Data processing step: Read calculation data according to the read-in decomposition descriptor, perform calculations on the calculation data through the four-dimensional variational assimilation system model, and output the calculated data according to the read-out decomposition descriptor; Among them, the mapping relationship configuration step is specifically: Divide the ocean model variable data into multiple data blocks, allocate a process to process each data block, read the boundary data of the adjacent process of the current process into the current process through the read-in decomposition descriptor, and directly write the real data of the current process into the storage file through the read-out decomposition descriptor; Calculate the offset by traversing the latitude direction through the first-layer loop, traverse the longitude direction through the second-layer loop, calculate the data index position according to the offset, and obtain the mapping relationship according to the data index position.

2. The data communication optimization method based on a regional ocean model according to claim 1, wherein The calculation model of the offset is: Among them, is the offset; is the latitude direction; is the offset of the ocean model variable in the latitude direction; is the global domain length in the longitude direction.

3. The data communication optimization method based on a regional ocean model according to claim 2, characterized in that The calculation model of the data index position is: Among them, is the data index position; is the longitude direction; is the offset of the ocean model variable in the longitude direction.

4. The data communication optimization method based on the regional ocean model according to claim 2, wherein The mapping relationship configuration step further includes: When configuring the data reading mapping relationship, the range in the latitude direction is traversed through the first-layer loop as from 1 to 1, and the range in the longitude direction is traversed through the second-layer loop as from 1 to 1; Among them, and 1、 1、 The calculation models of 1 are respectively: ； ； ； ； Among them, is the number of boundary-side lattice points.

5. The data communication optimization method based on a regional ocean model according to claim 3, wherein The mapping relationship configuration step further includes: When calculating the data readout mapping relationship, the range in the latitude direction is traversed through the first-layer loop from to , and the range in the longitude direction is traversed through the second-layer loop from to ; Among them, and 2、 2、 The calculation models of 2 are respectively: ； ； ； 。 6. The data communication optimization method based on the regional ocean model according to claim 5, wherein, The data processing step is specifically: In the outer loop of the four-dimensional variational assimilation system, load and read the calculation data according to the read-in decomposition descriptor through the non-linear model, and perform time stepping calculation according to the calculation data; In the inner loop of the four-dimensional variational assimilation system, load and read the calculation data according to the read-in decomposition descriptor through the adjoint model and the tangent linear model respectively, and perform time stepping calculation according to the calculation data.

7. The data communication optimization method based on the regional ocean model according to claim 6, characterized in that The data processing step further includes: Load a preset configuration file into the memory, output the calculated data according to the variable id and variable generation time configured in the configuration file through the read-out decomposition descriptor, and complete the outer loop and the inner loop according to the total time step and time step length configured in the configuration file.

8. An optimized data communication system based on a regional ocean model, characterized in that, It includes: Data acquisition module: Acquire ocean model variables in the regional ocean model, and define the read-in decomposition descriptor and the read-out decomposition descriptor of the data of the ocean model variables through the parallel input / output system; wherein, the read-in decomposition descriptor and the read-out decomposition descriptor define the storage location of the data of the ocean model variables in the file, the storage location of the data of the ocean model variables in the memory, and the chunking method of the data of the ocean model variables; Mapping relationship configuration module: Configure the data reading mapping relationship between the storage location of ocean model variable data in the storage file and the storage location in the memory for the data reading decomposition descriptor, and configure the data reading mapping relationship between the storage location of ocean model variable data in the storage file and the storage location in the memory for the data reading decomposition descriptor; Data processing module: Read the calculation data according to the data reading decomposition descriptor, perform calculations on the calculation data through the four-dimensional variational assimilation system model, and output the calculated data according to the data reading decomposition descriptor; Among them, the mapping relationship configuration module is specifically: Divide the ocean model variable data into multiple data blocks, allocate a process for each data block to process, read the boundary data of the adjacent process of the current process into the current process through the data reading decomposition descriptor, and directly write the real data of the current process into the storage file through the data reading decomposition descriptor; Calculate the offset by traversing the latitude direction through the first-layer loop, traverse the longitude direction through the second-layer loop, calculate the data index position according to the offset, and obtain the mapping relationship according to the data index position.

Images