Distributed storage method and system based on AI computing power analysis

Abstract

The invention discloses a distributed storage method and system based on AI computing power analysis. The method comprises the steps that heterogeneous storage resources are constructed in advance, and routing nodes are in communication connection with at least one storage node; obtaining computing power state data of routing nodes and storage nodes in the heterogeneous storage resources according to historical data of the heterogeneous storage resources, and converting computing power state information into computing power feature data of each routing node and each storage node; outputting computing power state data of each routing node and each storage node according to an AI computing power prediction model; and constructing a distributed storage task priority rule, storing the corresponding data packet to the corresponding storage node through the routing node according to the computing power state data of the routing node and the storage node and the distributed storage task priority rule, and updating the storage state data of the storage node and the routing node in real time. The adaptability of the distributed storage method to different storage task rules can be improved, and the stability of distributed storage is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of storage technology, and in particular to a distributed storage method and system based on AI computing power analysis. Background Art

[0002] At present, the traditional distributed storage method only performs sharded storage according to the storage capacity of different distributed nodes, and schedules and stores data files through load balancing. For example, the traditional HFDS system (distributed file system) includes named nodes and data nodes in the system, where the named node is the master node used to control the storage scheduling of the data node; however, the traditional HFDS system has high hardware requirements. Since the performance of a single master node has an upper limit, it needs to rely on multiple master nodes for processing when performing large-scale data processing and storage, thereby limiting the concurrent performance of the HFDS system. In addition, the traditional HFDS system has poor effects on low-latency business scenarios such as high-precision real-time queries, especially in high-concurrency real-time query scenarios. The traditional HFDS system is very likely to cause the master node to crash and cause the data node cluster connected to the master node to be unavailable. Therefore, the above-mentioned traditional distributed storage is prone to distributed storage cluster failures under high concurrency conditions due to the single node function and the problem of performance upper limit. Summary of the invention

[0003] One of the inventive purposes of the present invention is to provide a distributed storage method and system based on AI computing power analysis. The method and system construct an AI computing power prediction model of a distributed computing power cluster by acquiring dynamic computing power data consisting of routing nodes and storage nodes, converting the routing node-related data and storage node-related data into computing power feature data, and using the AI ​​computing power prediction model to predict the computing power values ​​of routing nodes and storage nodes of different levels, and performing distributed storage in combination with the business needs of the current storage task. The AI ​​computing power prediction model used in the present invention can better predict the computing power status of different routing nodes and storage nodes at different times, thereby meeting the distributed storage under specific storage business rules and improving the business adaptability of storage resource scheduling.

[0004] Another inventive object of the present invention is to provide a distributed storage method and system based on AI computing power analysis. The method and system also include using an AI computing power prediction model to perform heterogeneous computing power analysis of different computing power resources on different storage nodes, and judging the collaborative processing efficiency between different computing power resource devices of the corresponding storage nodes based on the results of the heterogeneous computing power analysis. That is, the present invention collects collaborative processing data between the different computing power resource devices, and converts the collaborative processing data into heterogeneous computing power feature data, and inputs the heterogeneous computing power feature data into the AI ​​computing power prediction model for training, so that the AI ​​computing power prediction model can predict the computing power status of the heterogeneous computing power analysis model of different storage nodes, so that the present invention can meet the requirements of heterogeneous computing power storage for distributed storage, and improve the heterogeneous processing efficiency of distributed storage tasks.

[0005] Another inventive object of the present invention is to provide a distributed storage method and system based on AI computing power analysis, wherein the method and system analyze and predict the total computing power of each routing node for different levels of routing nodes based on the AI ​​computing power prediction model, and perform distributed storage tasks for corresponding storage data packets at corresponding routing nodes according to the analysis and prediction results and the corresponding storage task allocation rules. The storage task rules include storage task rules with storage and reading timeliness as priority, storage task rules with persistence as priority, and storage task rules with structural form as priority, etc., thereby greatly improving the adaptability of the distributed storage method to different storage task rules and greatly improving the stability of distributed storage.

[0006] In order to achieve at least one of the above-mentioned invention objects, the present invention further provides a distributed storage method based on AI computing power analysis, the method comprising:

[0007] S01. Pre-constructing heterogeneous storage resources, wherein the heterogeneous storage resources include routing nodes and storage nodes at different levels, wherein the routing node is communicatively connected to at least one storage node or routing node;

[0008] S02. Obtain the computing power status data of each routing node and storage node in the heterogeneous storage resource according to the historical data of the heterogeneous storage resource, and convert the computing power status information into computing power characteristic data of each routing node and storage node;

[0009] S03. Use the computing power feature data to build an AI computing power prediction model, and output the computing power status data of each routing node and storage node in different time series according to the AI ​​computing power prediction model;

[0010] S04. Construct a distributed storage task priority rule, store the corresponding data packet in the corresponding storage node through the corresponding routing node according to the computing power status data of the corresponding routing node and storage node and the distributed storage task priority rule, and update the storage status data of the storage node and routing node in real time.

[0011] According to one of the preferred embodiments of the present invention, the heterogeneous storage resources include: at least one of a volatile memory, a non-volatile memory and an object memory as a storage module, and at least two of a CPU processor, a GPU processor, a DPU processor, a TPU processor, an FPGA module, a DPS processor, an ASIC chip and an NPU processor as processing modules, and the at least one storage module and the at least two processing modules are used to construct the heterogeneous storage resources, the heterogeneous storage resources are constructed in corresponding storage nodes, and a heterogeneous label of a corresponding type is configured for each heterogeneous storage resource, and the heterogeneous label is saved in a routing node maintained by the storage node for scheduling corresponding storage task priority rules.

[0012] According to another preferred embodiment of the present invention, the method for calculating the computing power status data of the storage node includes: obtaining the computing power status data of the storage node under each historical time series in the storage node according to the time series model construction rule, wherein the computing power status data includes: the remaining capacity H of the storage resource corresponding to the storage node, the floating point computing power Flops average value f of the processor of the corresponding storage node, and the read and write speed V of the storage node. 1 , the routing time t between the routing node and the corresponding storage node, the co-processor speed V between heterogeneous processors 2 The corresponding storage node computing power comparison value P is calculated using the following formula: P= , where the computing power comparison value P is the computing power status data, which is used to compare the computing power difference between the current storage node and all storage nodes, where V s1 is the calculated average read and write speed of all storage nodes, V s2 is the average collaborative processing speed among all storage node heterogeneous processors, f 0 It is the average floating point computing power (Flops) of all storage node processors.

[0013] According to another preferred embodiment of the present invention, the computing power status data of the storage node under the historical time series is normalized to the corresponding type of input data, and the normalized data is used as the cell state data of the LSTM model, and the data type and size of the input data packet are pre-processed as the input data of the LSTM model, and the output layer of the LSTM model outputs the predicted storage node computing power status value under the corresponding time series; and the mean square error is used as the loss function of the LSTM model to perform the regression task of the different computing power status data of the storage node, which is used to regress and predict the computing power status data of the corresponding storage node including the next time series.

[0014] According to another preferred embodiment of the present invention, the computing power status data of the routing node includes: the total storage capacity H of all storage nodes connected to the current routing node z , the average routing speed of the current routing node and all storage nodes and the sum of the parallel routing speed V z , the highest routing speed V of the current routing node and the connected storage node 3max and minimum routing speed V 3min , the routing speed V of the current routing node and the previous routing node 4 , the routing speed V of the current routing node and the next level routing node 5 , wherein the calculation method of the routing speed includes: obtaining a first timestamp when the current routing node receives a data packet, and when the data packet is received by a routing node or a storage node at a corresponding level, obtaining a second timestamp when received in real time, and obtaining the routing speed according to the first timestamp, the second timestamp and the data packet size.

[0015] According to another preferred embodiment of the present invention, the computing power status data of the routing node in the historical time series is normalized to the corresponding type of input data, and the normalized data is used as the cell state data of the LSTM model, and the data type and size of the input data packet are pre-processed as the input data of the LSTM model, the output layer of the LSTM model outputs the predicted routing computing power status value under the corresponding time series, and the mean square error is used as the loss function of the LSTM model to perform the regression task of the different computing power status data of the routing node, which is used to regress and predict the computing power status data of the corresponding routing node including the next time series.

[0016] According to another preferred embodiment of the present invention, the priority rules of the storage task include: time priority rule, persistence priority rule, processing mode priority rule, redundancy priority rule and structural form priority rule, wherein the time priority rule is used to establish storage interaction tasks in low-latency scenarios. Under the time priority rule, the total computing power time M of the routing node and the storage node is calculated: M=α V 1 +β V 2 +γ V 3max +λ V 4 +σ V 5 , where α, β, γ, λ and σ are different weight parameters. Under the time priority rule, when the packet size h is less than the total storage capacity H of the corresponding routing node and the corresponding storage node, z When the total computing power time is the largest, select max The predicted routing node is used as the target routing node to send the data packet, and the P with the largest computing power comparison value is selected. max Acts as a storage node and as a target storage node for distributed storage of data packets.

[0017] According to another preferred embodiment of the present invention, the data type in the data packet that needs to be stored and the type of processing method required for the data packet are determined, and a processor type tag is generated according to the data packet type and the processing method type, wherein the processor type tag is compared with the processor type tag of the corresponding storage node stored in the router node, and a routing node with the same processor type tag and satisfying the timeliness priority rule is selected to send the data packet.

[0018] In order to achieve at least one of the above-mentioned purposes of the invention, the present invention further provides a distributed storage system based on AI computing power analysis, and the system executes the above-mentioned distributed storage method based on AI computing power analysis.

[0019] The present invention further provides a computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the above-mentioned distributed storage method based on AI computing power analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Shown is a flow chart of a distributed storage method based on AI computing power analysis of the present invention. DETAILED DESCRIPTION

[0021] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles of the present invention defined in the following description can be applied to other embodiments, variations, improvements, equivalents, and other technical solutions that do not deviate from the spirit and scope of the present invention.

[0022] It is to be understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the element may be multiple, and the term "one" should not be understood as a limitation on the quantity.

[0023] Please combine Figure 1 The present invention discloses a distributed storage method and system based on AI computing power analysis, wherein the method comprises the following steps:

[0024] S01. Pre-constructing heterogeneous storage resources, wherein the heterogeneous storage resources include routing nodes and storage nodes at different levels, wherein the routing node is communicatively connected to at least one storage node or routing node;

[0025] S02. Obtain the computing power status data of each routing node and storage node in the heterogeneous storage resource according to the historical data of the heterogeneous storage resource, and convert the computing power status information into computing power characteristic data of each routing node and storage node;

[0026] S03. Use the computing power feature data to build an AI computing power prediction model, and output the computing power status data of each routing node and storage node in different time series according to the AI ​​computing power prediction model;

[0027] S04. Construct a distributed storage task priority rule, store the corresponding data packet in the corresponding storage node through the corresponding routing node according to the computing power status data of the corresponding routing node and storage node and the distributed storage task priority rule, and update the storage status data of the storage node and routing node in real time.

[0028] In the present invention, it is necessary to pre-build a distributed storage node cluster, wherein the distributed storage node cluster includes routing nodes and storage nodes, and configure corresponding resource devices in the corresponding routing nodes and storage nodes. The resource devices of the storage nodes in the present invention include memory resources and processor resources, wherein the storage nodes can be configured to include volatile memory or non-volatile memory. The volatile memory may include but is not limited to DRAM (dynamic random access memory) and SRAM (static random access memory), and the volatile memory is generally used for memory storage and is often used for high-performance computing. The non-volatile memory includes but is not limited to Flash Memory (flash memory), SSD (solid state drive), MRAM (magnetic memory) and EPROM (erasable memory), etc. An appropriate number of the above-mentioned volatile memory and non-volatile memory are pre-configured in the corresponding storage node, so that the storage node cluster has good adaptability to multiple storage services. The processor resources in the corresponding storage node can be configured to include but are not limited to CPU processors, GPU processors, DPU processors, TPU processors, FPGA modules, DPS processors, ASIC chips and NPU processors, etc. The above-mentioned different types of processors can solve different data packet service processor problems. The above-mentioned processor and memory can be configured at the network edge node, and because different businesses require different types of processor resources, the present invention can pre-set heterogeneous resource configurations including at least the above two types of processors and at least one type of memory as corresponding storage nodes, so that the distributed storage nodes in the present invention have the ability to improve storage business efficiency and compatibility with different storage business rules. The present invention uses an AI model with a time series to construct a prediction model for the computing power status of the storage nodes and routing nodes, and performs distributed storage of corresponding data based on the setting of relevant storage business rules and the computing power prediction results of different storage nodes and routing nodes. The corresponding storage data packets in the present invention can be stored in a diversified manner based on different business rules, thereby adapting to business scenarios where different storage resources are called.

[0029] Specifically, the routing nodes can be configured into different levels, for example, the routing nodes can be constructed into a tree structure, each routing node is connected to at least one storage node or routing node in communication, and different routing nodes are configured including routing processors, etc. In another feasible embodiment of the present invention, if a storage cluster with a topological network structure is constructed, corresponding storage resources can be configured on the corresponding routing nodes. So that the routing nodes can be used as storage nodes, and the corresponding storage nodes can be used as routing nodes under certain circumstances. It should be noted that the computing power status data of the routing nodes and storage nodes in the present invention are different, wherein the computing power of the storage node is mainly determined by the overall performance of the computing resources of the current storage node, the read and write performance of the memory, the storage capacity and the routing time, and the overall computing power of the routing node is the sum of the computing power of the storage nodes of all branches, and the computing power status data calculation of the routing node in the present invention also includes the routing time between routing nodes of different levels, providing reference data for the data packet routing planning of different routing nodes. It is worth mentioning that the present invention needs to use the historical data of the distributed storage node cluster to train the AI ​​model, and use the historical computing power status data of each storage node and routing node in the node cluster as training data to train the AI ​​model; wherein the present invention preferably uses a long short-term memory network (LSTM) as an AI model carrying a time series to train the historical computing power status data of the storage node routing node, and obtain a computing power status prediction model for the corresponding storage node and routing node. The present invention performs distributed routing storage of corresponding data packets according to the computing power status data of each routing node and each storage node under the corresponding time series predicted by the AI ​​model and the storage task priority rules through pre-configured storage task priority rules. Thereby, the distributed storage is adapted to various business scenarios.

[0030] The long short-term memory network (LSTM) includes a cell state C t , input gate i t 、Forget gate t , output gate o t , and integrate the input gate data and the forget gate data input by element-by-element multiplication to update the cell state. In the present invention, the computing power state data of the routing node and the storage node are normalized to the corresponding type of input data, and the normalized data is used as the cell state data C of the long short-term memory network (LSTM) model. t , and the data including but not limited to the data packet type and size obtained after the corresponding storage node performs data preprocessing on the input data packet in the corresponding time series t as the input gate i t data, and combined with the forget gate f t Update the cell state C t+1The long short-term memory network (LSTM) model uses the mean square error as the loss function, calculates the mean square error value of the different computing power status data predicted by each storage node and routing node, and performs the regression task of the computing power status data of each storage node and routing node, which is used to predict the computing power status value of the corresponding storage node and routing node computing power status data under a specific time series. The training method and optimized training method of the long short-term memory network (LSTM) model are prior arts, and the present invention will not be described in detail.

[0031] It is worth mentioning that the present invention needs to perform routing storage of corresponding data packets according to the computing power state data predicted by the routing node and the actual storage task priority rules. For example, the storage task priority rules in the present invention include but are not limited to: timeliness priority rules, persistence priority rules, processing method priority rules, redundancy priority rules and structural form priority rules, wherein the timeliness priority rules are used for low-latency scene storage tasks, such as real-time data analysis scenes, real-time interactive autonomous driving scenes and real-time financial transaction scenes, etc. The above-mentioned low-latency scene storage tasks can use DRAM (dynamic random access memory) as a real-time cache, such as Redis, etc. The persistence priority rule is applicable to data persistence operations, such as redundant storage of videos through RAID (independent disk redundant array), which can automatically recover when some disks fail, thereby improving the security of data storage. The processing method priority is set according to the processing rules for the data packet type, such as some processing methods require the joint processing operation of heterogeneous processing resources of CPU and GPU, such as parallel analysis and calculation of fluid mechanics analysis using GPU, and also need to use CPU to set parameter boundary conditions and other pre-processing of data; therefore, it needs to be allocated to a storage node with specific heterogeneous processing resources. Similarly, for structured priority rules, storage nodes with structured memory need to be given priority. In the present invention, it is necessary to label the heterogeneous storage resources and processing resource types in each storage node, obtain all labels of the heterogeneous storage resources and processing resources in each storage node, and pre-store the labels in the routing node. By performing label comparison in the routing node, it can be used for data packet routing operations of storage nodes that require heterogeneous processing resources.

[0032] The method for obtaining and processing the computing power status data of the storage node comprises: obtaining the computing power status data of the storage node in each historical time series of the storage node according to the time series model construction rule, wherein the computing power status data comprises: the remaining capacity H of the storage resource corresponding to the storage node, the average floating point computing power Flops of the processor corresponding to the storage node, and the storage node memory read and write speed V 1, the routing time t between the routing node and the corresponding storage node, the co-processor speed V between heterogeneous processors 2 The corresponding storage node computing power comparison value P is calculated using the following formula: P= , where the computing power comparison value P is the computing power status data, which is used to compare the computing power difference between the current storage node and all storage nodes, where V s1 is the calculated average read and write speed of all storage nodes, V s2 is the average collaborative processing speed among all storage node heterogeneous processors, f 0 is the average value of the floating point computing power Flops of all storage node processors. It should be noted that the above computing power comparison value is set in the present invention for comparing the computing power comparison in the storage nodes, wherein the computing power comparison of the storage nodes is compatible with the storage capacity and processing speed parameter considerations of the storage nodes themselves. Of course, in other feasible embodiments of the present invention, it is also possible to select parameters that only consider storage capacity or processing speed as data routing storage parameters according to specific storage business rules.

[0033] The computing power status data of the routing node includes: the total storage capacity H of all storage nodes connected to the current routing node z , the average routing speed of the current routing node and all storage nodes and the sum of the parallel routing speed V z , the highest routing speed V of the current routing node and the connected storage node 3max and minimum routing speed V 3min , the routing speed V of the current routing node and the previous routing node 4 , the routing speed V of the current routing node and the next level routing node 5 , wherein the calculation method of the routing speed includes: obtaining a first timestamp when the current routing node receives a data packet, and when the data packet is received by a routing node or a storage node at a corresponding level, obtaining a second timestamp when received in real time, and obtaining the routing speed according to the first timestamp, the second timestamp and the data packet size.

[0034] In one of the preferred embodiments of the present invention, in order to better explain the different storage task priority rules, the present invention takes the time-effective storage priority rule as an example: Under the time-effective priority rule, the total computing power time M of the routing node and the corresponding storage node is calculated: M=α V 1 +β V 2 +γ V 3max +λ V 4 +σ V5 , where α, β, γ, λ and σ are different weight parameters. Under the time priority rule, when the packet size h is less than the total storage capacity H of the corresponding routing node and the corresponding storage node, z When the total computing power time is the largest, select max The predicted routing node is used as the target routing node to send the data packet, and the P with the largest computing power comparison value is selected. max As a storage node, as a target storage node, data packets are distributedly stored. In another feasible embodiment of the present invention, the specific parameters of the total computing power time M and the computing power comparison value P of the routing node and the corresponding storage node can be taken into account for weighted processing to obtain the corresponding corresponding routing node computing power reference data for routing storage processing of the corresponding data packet. The present invention does not make any specific restrictions on this.

[0035] The embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. The embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part, and / or installed from a removable medium. When the computer program is executed by the central processing unit (CPU), the above functions defined in the method of the present application are executed. It should be noted that the computer-readable medium mentioned above in the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of an electrical, magnetic, optical, electromagnetic, infrared segment, or semiconductor, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more wire segments, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, electrical wire, optical cable, RF, etc., or any suitable combination of the foregoing.

[0036] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0037] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functional and structural principles of the present invention have been demonstrated and explained in the embodiments. Without departing from the principles, the implementation methods of the present invention may be deformed or modified in any way.

Claims

1. A distributed storage method based on AI computing power analysis, characterized in that: The method comprises: S01. Pre-constructing heterogeneous storage resources, wherein the heterogeneous storage resources include routing nodes and storage nodes at different levels, wherein the routing node is communicatively connected to at least one storage node or routing node; S02. Obtain the computing power status data of each routing node and storage node in the heterogeneous storage resource according to the historical data of the heterogeneous storage resource, and convert the computing power status information into computing power characteristic data of each routing node and storage node; S03. Use the computing power feature data to build an AI computing power prediction model, and output the computing power status data of each routing node and storage node in different time series according to the AI ​​computing power prediction model; S04. Construct a distributed storage task priority rule, store the corresponding data packet in the corresponding storage node through the corresponding routing node according to the computing power status data of the corresponding routing node and storage node and the distributed storage task priority rule, and update the storage status data of the storage node and routing node in real time.

2. According to the distributed storage method based on AI computing power analysis according to claim 1, it is characterized in that: The heterogeneous storage resources include: at least one of a volatile memory, a non-volatile memory and an object memory as a storage module, and at least two of a CPU processor, a GPU processor, a DPU processor, a TPU processor, an FPGA module, a DPS processor, an ASIC chip and an NPU processor as processing modules, and the at least one storage module and the at least two processing modules are used to construct the heterogeneous storage resources. The heterogeneous storage resources are constructed in corresponding storage nodes, and a corresponding type of heterogeneous label is configured for each heterogeneous storage resource, and the heterogeneous label is saved in a routing node maintained by the storage node for scheduling corresponding storage task priority rules.

3. According to the distributed storage method based on AI computing power analysis according to claim 1, it is characterized in that: The method for calculating the computing power status data of the storage node includes: obtaining the computing power status data of the storage node in each historical time series in the storage node according to the time series model construction rule, wherein the computing power status data includes: the remaining capacity H of the storage resource of the corresponding storage node, the floating point computing power Flops average value f of the processor of the corresponding storage node, the read and write speed V1 of the storage node, the routing time t of the routing node and the corresponding storage node, and the co-processor speed V2 between heterogeneous processors, and calculating the computing power comparison value P of the corresponding storage node through the following formula: P= , where the computing power comparison value P is the computing power status data, which is used to compare the computing power difference between the current storage node and all storage nodes, where V s1 is the calculated average read and write speed of all storage nodes, V s2 is the average value of the collaborative processing speed among heterogeneous processors of all storage nodes, and f0 is the average value of the floating-point computing power (Flops) of processors of all storage nodes.

4. According to claim 3, a distributed storage method based on AI computing power analysis is characterized in that: The computing power status data of the storage node in the historical time series is normalized according to the corresponding type of input data, and the normalized data is used as the cell state data of the LSTM model, and the data type and size of the input data packet are pre-processed as the input data of the LSTM model, and the output layer of the LSTM model outputs the predicted storage node computing power status value in the corresponding time series; and the regression task of the different computing power status data of the storage node is performed by using the mean square error as the loss function of the LSTM model, which is used to regress and predict the computing power status data of the corresponding storage node including the next time series.

5. A distributed storage method based on AI computing power analysis according to claim 4, characterized in that: The computing power status data of the routing node includes: the total storage capacity H of all storage nodes connected to the current routing node z , the average routing speed of the current routing node and all storage nodes and the sum of the parallel routing speed V z , the highest routing speed V of the current routing node and the connected storage node 3max and minimum routing speed V 3min , the routing speed V4 of the current routing node and the previous level routing node, the routing speed V5 of the current routing node and the next level routing node, wherein the calculation method of the routing speed includes: obtaining a first timestamp when the current routing node receives a data packet, when the data packet is received by a routing node or a storage node of the corresponding level, obtaining a second timestamp at the time of reception in real time, and obtaining the routing speed according to the first timestamp, the second timestamp and the data packet size.

6. A distributed storage method based on AI computing power analysis according to claim 4, characterized in that: The computing power status data of the routing node in the historical time series is normalized to the corresponding type of input data, and the normalized data is used as the cell state data of the LSTM model, and the data type and size of the input data packet are pre-processed as the input data of the LSTM model. The output layer of the LSTM model outputs the predicted routing computing power status value under the corresponding time series, and the mean square error is used as the loss function of the LSTM model to perform the regression task of the different computing power status data of the routing node, which is used to regress and predict the computing power status data of the corresponding routing node including the next time series.

7. The distributed storage method based on AI computing power analysis according to claim 5 is characterized in that: The priority rules of the storage task include: time priority rule, persistence priority rule, processing method priority rule, redundancy priority rule and structural form priority rule, wherein the time priority rule is used to establish storage interaction tasks in low-latency scenarios. Under the time priority rule, the total computing power time M of the routing node and the storage node is calculated: M=α V1+β V2+γ V 3max +λ V4+σ V5, where α, β, γ, λ and σ are different weight parameters. Under the time priority rule, when the packet size h is less than the total storage capacity H of the corresponding storage node of the corresponding routing node, z When the total computing power time is the largest, select max The predicted routing node is used as the target routing node to send data packets, and the P with the largest computing power comparison value is selected. max Acts as a storage node and as a target storage node for distributed storage of data packets.

8. The distributed storage method based on AI computing power analysis according to claim 7 is characterized in that: Determine the type of data in the data packet that needs to be stored, and the type of processing method required for the data packet, generate a processor type tag according to the data packet type and the processing method type, compare the processor type tag with the processor type tag of the corresponding storage node stored in the router node, and select a routing node with the same processor type tag that meets the time priority rule to send the data packet.

9. A distributed storage system based on AI computing power analysis, characterized in that: The system executes a distributed storage method based on AI computing power analysis as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement a distributed storage method based on AI computing power analysis as described in any one of claims 1-8.

Images