OpenSAI intelligent connection network system and computing network matching method and system

Abstract

The invention relates to the technical field of network system service, and particularly discloses an OpenSAI intelligent connection network system and a computing network matching method and system. According to the invention, the resource monitoring module monitors the states of the network nodes unloaded into the computing node cluster by the plurality of application tasks to obtain the plurality of node performance monitoring values, and sequentially judges whether the plurality of node performance monitoring values meet the preset monitoring value, if not, the application tasks are unloaded into the computing node cluster, and if not, the application tasks are unloaded into the computing node cluster. If yes, a first resource scheduling instruction of the corresponding required resource is generated for the node which does not meet the preset monitoring value, and after receiving the first resource scheduling instruction, the resource control module dynamically schedules the calculation and storage network resource required by the corresponding node according to the first resource scheduling instruction. And meanwhile, the task can be executed under a proper condition, so that the overall performance and the response speed of the system are improved, and the method is suitable for long-term edge task scheduling.

Description

Technical Field

[0001] The present invention relates to the technical field of network system services, and in particular to an OpenSAI intelligent network system and a computing network matching method and system. Background Art

[0002] OpenSAI is a system that emphasizes openness, intelligence, and networking. Specifically, it is an open platform or framework for smart devices or network systems. It aims to provide a series of standardized interfaces and services to facilitate interconnection between different devices and services. Currently, artificial intelligence applications are mainly deployed in the cloud. In the process of artificial intelligence applications, it is usually necessary to deploy lightweight machines on edge computing platforms, and to use real-time IoT data more intelligently under the application of lightweight machines. At the same time, with the development of open source container orchestration engines and cloud-native technologies, edge computing platforms require more resources, and edge computing platforms need more cluster resource management and application lightweight deployment capabilities.

[0003] As big data real-time computing and machine learning tasks consume a large amount of edge cluster resources, effective scheduling is required. Existing research is mostly based on deep learning and heuristic algorithm design scheduling, but the number of training times is too many, which is not suitable for edge platforms, and the algorithms are mostly offline, which is not suitable for long-term edge task scheduling. Therefore, an OpenSAI intelligent network system and a computing network matching method and system are needed to solve the problem of not being suitable for long-term edge task scheduling. Summary of the invention

[0004] The purpose of the present invention is to provide an OpenSAI intelligent network system and a computing network matching method and system to solve the technical problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: An OpenSAI intelligent network system, comprising: an application task module, a cluster management module, a computing distribution module, an adaptation optimization module, a resource monitoring module and a resource control module; The application task module is used to obtain multiple application tasks, sort the multiple application tasks to obtain a task sorting list, and perform interface parameter resource configuration on the multiple application tasks in sequence according to the task sorting list to obtain multiple application task configuration resources, and send the multiple application task configuration resources to the computing cluster management module; The cluster management module is used to obtain the current node status and configure resources to form a computing node cluster according to the current node status and the multiple application tasks; The computing distributed module is used to offload multiple application tasks to the network nodes in the computing node cluster for execution, thereby obtaining multiple execution tasks; The adaptation optimization module is used to dynamically configure the plurality of execution tasks based on the resources provided by the elastic base, wherein the dynamic configuration includes application task configuration, application task and network node configuration and network resource scheduling configuration; The resource monitoring module is used to monitor the status of the network nodes in the computing node cluster to which multiple application tasks are offloaded, obtain multiple node performance monitoring values, and sequentially determine whether the multiple node performance monitoring values ​​meet the preset monitoring values; If not, generating a first resource scheduling instruction of the corresponding required resource for the node that does not meet the preset monitoring value; The resource control module is used to receive a first resource scheduling instruction and dynamically schedule the computing and storage network resources required by the corresponding node according to the first resource scheduling instruction.

[0006] Preferably, the adaptation optimization module is also used to simultaneously optimize the application task configuration and network bandwidth and network computing resources, obtain task-network bidirectional synchronization optimization information, and pair the application tasks and network nodes based on the task-network bidirectional synchronization optimization information.

[0007] The present invention also provides a computing network matching method for executing the OpenSAI intelligent network system according to any one of claims 1 to 2, comprising: Acquire multiple application tasks, sort the multiple application tasks to obtain a task sorting list, and sequentially configure interface parameter resources for the multiple application tasks according to the task sorting list to obtain multiple application task configuration resources; Obtaining the current node status, and configuring resources to form a computing node cluster according to the current node status and the multiple application tasks; Offloading multiple application tasks to network nodes in a computing node cluster for execution, thereby obtaining multiple execution tasks; Dynamically configuring the plurality of execution tasks based on resources provided by the elastic base, wherein the dynamic configuration includes application task configuration, application task and network node configuration, and network resource scheduling configuration; Monitor the status of the network nodes in the computing node cluster to which the multiple application tasks are offloaded, obtain multiple node performance monitoring values, and determine in turn whether the multiple node performance monitoring values ​​meet the preset monitoring values; If not, generating a first resource scheduling instruction of the corresponding required resource for the node that does not meet the preset monitoring value; The computing and storage network resources required by the corresponding node are dynamically scheduled according to the first resource scheduling instruction.

[0008] Preferably, the step of sequentially configuring interface parameter resources for the plurality of application tasks according to the task sorting list to obtain a plurality of application task configuration resources further includes: Acquire a first CPU occupied resource amount, a first RAM occupied resource amount, a first GPU occupied resource amount, and a first bandwidth occupied resource amount of the application task according to the application task configuration resources; Calculating a first CPU occupancy ratio according to the first CPU occupied resource amount and a preset CPU occupied resource amount; Calculating a first RAM occupancy ratio according to the first RAM occupied resource amount and a preset RAM occupied resource amount; Calculating a first GPU occupancy ratio according to the first GPU occupied resource amount and the preset GPU occupied resource amount; Calculating a first bandwidth occupancy ratio according to the first bandwidth occupied resource amount and a preset bandwidth occupied resource amount; The resource occupancy score of the application task is calculated according to the first CPU occupancy ratio, the first RAM occupancy ratio, the first GPU occupancy ratio, and the first bandwidth occupancy ratio, wherein the calculation formula is: ; in, Indicates the resource usage score of the application task. Indicates the first CPU usage ratio. Indicates the first RAM occupancy ratio, Indicates the first GPU occupancy rate, Indicates the first bandwidth occupancy ratio; According to the above steps, multiple application task resource occupancy scores corresponding to the multiple application task configuration resources are calculated in sequence.

[0009] Preferably, the step of monitoring the status of the network nodes in the computing node cluster to offload the multiple application tasks to obtain multiple node performance monitoring values ​​further includes: Obtaining a second CPU occupied resource amount, a second RAM occupied resource amount, a second GPU occupied resource amount, and a second bandwidth occupied resource amount of the node according to the node performance monitoring value; Calculate the second CPU occupancy ratio according to the second CPU occupied resource amount and the preset CPU occupied resource amount; Calculate a second RAM occupancy ratio according to the second RAM occupied resource amount and a preset RAM occupied resource amount; Calculating the second GPU occupation ratio according to the second GPU occupation amount and the preset GPU occupation amount; Calculating a second bandwidth occupation ratio according to the second bandwidth occupied resource amount and a preset bandwidth occupied resource amount; The node resource occupancy score is calculated according to the second CPU occupancy ratio, the second RAM occupancy ratio, and the second bandwidth occupancy ratio, wherein the calculation formula is: ; in, Indicates the node resource usage score, Indicates the second CPU usage ratio. Indicates the second RAM occupancy ratio, Indicates the first GPU occupancy rate, represents the second bandwidth occupancy ratio; According to the above calculation steps, multiple node resource occupancy scores corresponding to the multiple node performance monitoring values ​​are calculated in sequence.

[0010] Preferably, the step of dynamically scheduling the computing and storage network resources required by the corresponding node according to the first resource scheduling instruction further includes: The resource occupancy score of the application task is traversed and matched with the resource occupancy scores of multiple nodes based on the similarity model to obtain a resource similarity value, wherein the function of the similarity model is: ; in, Indicates the resource similarity value, Indicates the resource usage score of the application task. It represents the node resource usage score, n represents the sequence number of the node resource usage score, n=1, 2, 3...n.

[0011] The present application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0012] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0013] The beneficial effects of the present application are as follows: the present invention first obtains multiple application tasks through the application task module, and prepares the priority and configuration parameters of the application tasks. Then the cluster management module is used to obtain the current node status and form a computing node cluster. At the same time, multiple application tasks are unloaded to the network nodes in the computing node cluster for execution to obtain multiple execution tasks. Then the adaptation optimization module dynamically configures the multiple execution tasks based on the resources provided by the elastic base. Finally, the resource monitoring module is used to monitor the status of the network nodes in the computing node cluster to which the multiple application tasks are unloaded, obtain multiple node performance monitoring values, and judge in turn whether the multiple node performance monitoring values ​​meet the preset monitoring values. If not, the first resource scheduling instruction of the corresponding required resource is generated for the node that does not meet the preset monitoring value. After receiving the first resource scheduling instruction, the resource control module dynamically schedules the computing storage network resources required by the corresponding node according to the first resource scheduling instruction. In this way, the system can dynamically match the application task configuration resources and the node resources, and then make the optimal resource allocation decision according to the real-time data. This method can ensure that resources are efficiently utilized, and at the same time, it can also ensure that the tasks are executed under appropriate conditions, thereby improving the overall performance and response speed of the system, and is suitable for the scheduling of long-term edge tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The figure is a schematic diagram of a method flow of an embodiment of the present application.

[0015] Figure 2 A schematic diagram of the system structure of an embodiment of the present application.

[0016] Figure 3 A schematic diagram of the internal structure of a computer device according to an embodiment of the present application.

[0017] Figure 4 This is a schematic diagram illustrating an example of an embodiment of the present application.

[0018] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0019] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0020] like Figure 1-3 As shown, the present application provides an OpenSAI intelligent network system, including: an application task module, a cluster management module, a computing distribution module, an adaptation optimization module, a resource monitoring module and a resource control module; The application task module is used to obtain multiple application tasks, sort the multiple application tasks to obtain a task sorting list, and perform interface parameter resource configuration on the multiple application tasks in sequence according to the task sorting list to obtain multiple application task configuration resources, and send the multiple application task configuration resources to the computing cluster management module; The cluster management module is used to obtain the current node status and configure resources to form a computing node cluster according to the current node status and the multiple application tasks; The computing distributed module is used to offload multiple application tasks to the network nodes in the computing node cluster for execution, thereby obtaining multiple execution tasks; The adaptation optimization module is used to dynamically configure the plurality of execution tasks based on the resources provided by the elastic base, wherein the dynamic configuration includes application task configuration, application task and network node configuration and network resource scheduling configuration; The resource monitoring module is used to monitor the status of the network nodes in the computing node cluster to which multiple application tasks are offloaded, obtain multiple node performance monitoring values, and sequentially determine whether the multiple node performance monitoring values ​​meet the preset monitoring values; If not, generating a first resource scheduling instruction of the corresponding required resource for the node that does not meet the preset monitoring value; The resource control module is used to receive a first resource scheduling instruction and dynamically schedule the computing and storage network resources required by the corresponding node according to the first resource scheduling instruction.

[0021] The present invention obtains multiple application tasks through an application task module, sorts the multiple application tasks to obtain a task sorting list, and sequentially configures interface parameter resources for the multiple application tasks according to the task sorting list to obtain multiple application task configuration resources, and sends the multiple application task configuration resources to a computing cluster management module, so that a corresponding task instance can be quickly created according to each application task, and parameters such as the priority of the task and the configuration interface can be set to provide a basis for the calculation of subsequent application tasks, and then the current node status is obtained through the cluster management module, and a computing node cluster is formed according to the current node status and the multiple application task configuration resources, so that the cluster management module collects the status information of all currently available nodes (such as CPU usage, memory occupancy, network status, etc.), and then the computing resource availability can be evaluated according to the node status to prepare for task allocation, and then the computing distributed module is used to unload the multiple application tasks to the network nodes in the computing node cluster for execution, and multiple execution tasks are obtained. At the same time, each node works in coordination through a communication module, wherein the computing distributed module is also used to analyze the complexity of the task, determine whether distributed computing is required, and allocate the execution tasks. Specific parameters and resource requirements are given to each node, and then the adaptation optimization module dynamically configures the multiple execution tasks based on the resources provided by the elastic base, wherein the dynamic configuration includes application task configuration, application task and network node configuration and network resource scheduling configuration, so that the adaptation optimization module can optimize the application and the network at the same time, thereby ensuring the real-time and reliability of task execution, and finally, the resource monitoring module monitors the status of the network nodes in the computing node cluster to which the multiple application tasks are unloaded, obtains multiple node performance monitoring values, and judges in turn whether the multiple node performance monitoring values ​​meet the preset monitoring values, if not, then the first resource scheduling instruction of the corresponding required resource is generated for the node that does not meet the preset monitoring value, and the resource control module is used to receive the first resource scheduling instruction, and dynamically schedule the computing storage network resources required by the corresponding node according to the first resource scheduling instruction, so that the system can dynamically match the application task configuration resources and the node resources, and then make the best resource allocation decision according to the real-time data, this method can ensure that resources are efficiently utilized, and at the same time can also ensure that the task is executed under appropriate conditions, thereby improving the overall performance and response speed of the system, and is suitable for the scheduling of long-term edge tasks.

[0022] In one embodiment, the adaptation optimization module is also used to simultaneously optimize the application task configuration and network bandwidth and network computing resources, obtain task-network bidirectional synchronization optimization information, and pair application tasks and network nodes based on the task-network bidirectional synchronization optimization information.

[0023] The present invention optimizes the application task configuration and the network bandwidth and network computing power resources simultaneously through the adaptation optimization module to obtain task-network bidirectional synchronous optimization information, so that the execution environment can be optimized, thereby ensuring the real-time and reliability of task execution, and based on the task-network bidirectional synchronous optimization information, the application resource configuration and multiple node performance monitoring values ​​are traversed and paired, so that after the application resource configuration and multiple node performance monitoring values ​​are traversed and paired, the application can be calculated and run on the most suitable node.

[0024] The present invention also provides a computing network matching method for executing the OpenSAI intelligent network system according to any one of claims 1 to 2, comprising: S1. Acquire multiple application tasks, sort the multiple application tasks to obtain a task sorting list, and sequentially perform interface parameter resource configuration on the multiple application tasks according to the task sorting list to obtain multiple application task configuration resources; S2. Obtain the current node status, and configure resources to form a computing node cluster according to the current node status and the multiple application tasks; S3, offloading multiple application tasks to network nodes in the computing node cluster for execution, thereby obtaining multiple execution tasks; S4. Dynamically configure the plurality of execution tasks based on resources provided by the elastic base, wherein the dynamic configuration includes application task configuration, application task and network node configuration, and network resource scheduling configuration; S5. Monitor the status of the network nodes in the computing node cluster to which the multiple application tasks are offloaded, obtain multiple node performance monitoring values, and determine in turn whether the multiple node performance monitoring values ​​meet the preset monitoring values; If not, generating a first resource scheduling instruction of the corresponding required resource for the node that does not meet the preset monitoring value; S6. Dynamically schedule the computing and storage network resources required by the corresponding node according to the first resource scheduling instruction.

[0025] As described in the above steps S1-S6, the present invention first obtains multiple application tasks, sorts the multiple application tasks, obtains a task sorting list, and sequentially configures interface parameter resources for the multiple application tasks according to the task sorting list to obtain multiple application task configuration resources, then obtains the current node state, and forms a computing node cluster according to the current node state and the multiple application task configuration resources, then unloads the multiple application tasks to the network nodes in the computing node cluster for execution to obtain multiple execution tasks, and then dynamically configures the multiple execution tasks based on the resources provided by the elastic base, wherein the dynamic configuration includes application task configuration, application task and network node configuration and network resource scheduling configuration, and finally, monitors the status of the network nodes in the computing node cluster to which the multiple application tasks are unloaded to obtain multiple node performance monitoring values, and sequentially determines whether the multiple node performance monitoring values ​​meet the preset monitoring values, so that the system can dynamically match the application task configuration resources and the node resources, and then make the optimal resource allocation decision according to the real-time data, thereby improving the overall performance and response speed of the system, and is suitable for the scheduling of long-term edge tasks, such as the actual application of this method (such as Figure 4 shown), as mentioned above Figure 4 Applications are used for illustration purposes only and are not intended to be exclusive.

[0026] In one embodiment, the step S1 of sequentially configuring interface parameter resources for the plurality of application tasks according to the task sorting list to obtain a plurality of application task configuration resources further includes: S101, acquiring a first CPU occupied resource amount, a first RAM occupied resource amount, a first GPU occupied resource amount, and a first bandwidth occupied resource amount of an application task according to the application task configuration resources; S102, calculating a first CPU occupancy ratio according to the first CPU occupied resource amount and a preset CPU occupied resource amount; S103, calculating a first RAM occupancy ratio according to the first RAM occupied resource amount and a preset RAM occupied resource amount; S104, calculating a first GPU occupancy ratio according to the first GPU occupied resource amount and a preset GPU occupied resource amount; S105: Calculate a first bandwidth occupancy ratio according to the first bandwidth occupied resource amount and a preset bandwidth occupied resource amount; S106. Calculate the resource occupancy score of the application task according to the first CPU occupancy ratio, the first RAM occupancy ratio, the first GPU occupancy ratio, and the first bandwidth occupancy ratio, wherein the calculation formula is: ; in, Indicates the resource usage score of the application task. Indicates the first CPU usage ratio. Indicates the first RAM occupancy ratio, Indicates the first GPU occupancy rate, Indicates the first bandwidth occupancy ratio; S107. Calculate, in sequence according to the above steps, a plurality of application task resource occupancy scores corresponding to the plurality of application task configuration resources.

[0027] As described in the above steps S102-S107, the present invention first obtains the first CPU occupied resource amount, the first RAM occupied resource amount, the first GPU occupied resource amount and the first bandwidth occupied resource amount of the application task according to the application task configuration resources, and then calculates the first CPU occupancy ratio according to the first CPU occupied resource amount and the preset CPU occupied resource amount. By calculating the first CPU occupancy ratio of the device, that is, the ratio of the actually used CPU resources to the preset CPU resources, the CPU resource usage of the device can be understood, so as to further analyze and adjust the resource allocation strategy, and then calculates the first RAM occupancy ratio according to the first RAM occupied resource amount and the preset RAM occupied resource amount, and obtains the RAM resource usage of the device in order to further analyze and adjust the resource allocation strategy, and then calculates the first bandwidth occupancy ratio according to the first bandwidth occupied resource amount and the preset bandwidth occupied resource amount. Finally, the application task resource occupancy score is calculated according to the first CPU occupancy ratio, the first RAM occupancy ratio and the first bandwidth occupancy ratio, and then according to the above The step sequentially calculates the resource occupancy scores of multiple application tasks corresponding to the application task configuration resources. For example, the first CPU resource occupancy is 1000, the preset CPU resource occupancy is 2000, the first GPU resource occupancy is 1000, the preset GPU resource occupancy is 2000, the first RAM resource occupancy is 512, the preset RAM resource occupancy is 1024, the first bandwidth occupancy ratio is 100, the preset bandwidth occupancy is 200, and the first CPU occupancy ratio = 1000 / 2000 = 0.5, the first RAM occupancy ratio = 512 / 1024 = 0.5, the first GPU occupancy ratio = 1000 / 2000 = 0.5, the first bandwidth occupancy ratio = 100 / 200 = 0.5, the application task resource occupancy score = (0.5+0.5+0.5+0.5) / 4 = 0.5. In this way, the resource usage of the device is evaluated by the application task resource occupancy score, which is convenient for managers to optimize resources and make decisions, and provides an important basis for resource scheduling. The above examples are only for illustration and are not the only reference.

[0028] In one embodiment, the step S5 of monitoring the status of the network nodes in the computing node cluster to obtain multiple node performance monitoring values ​​according to the multiple application tasks being offloaded to the network nodes further includes: S501, obtaining a second CPU occupied resource amount, a second RAM occupied resource amount, a second GPU occupied resource amount, and a second bandwidth occupied resource amount of a node according to the node performance monitoring value; S502, calculating a second CPU occupancy ratio according to the resource occupancy amount of the second CPU and the resource occupancy amount of a preset CPU; S503, calculating a second RAM occupancy ratio according to the second RAM occupied resource amount and a preset RAM occupied resource amount; S504, calculating the second GPU occupancy ratio according to the second GPU occupied resource amount and the preset GPU occupied resource amount; S505: Calculate a second bandwidth occupancy ratio according to the second bandwidth occupied resource amount and a preset bandwidth occupied resource amount; S506. Calculate the node resource occupancy score according to the second CPU occupancy ratio, the second RAM occupancy ratio, and the second bandwidth occupancy ratio, wherein the calculation formula is: ; in, Indicates the node resource usage score, Indicates the second CPU usage ratio. Indicates the second RAM occupancy ratio, Indicates the first GPU occupancy rate, represents the second bandwidth occupancy ratio; S507: Calculate multiple node resource occupancy scores corresponding to multiple node performance monitoring values ​​in sequence according to the above calculation steps.

[0029] As described in the above steps S501-S507, the present invention first obtains the second CPU occupied resource amount, the second RAM occupied resource amount, the second GPU occupied resource amount and the second bandwidth occupied resource amount of the node according to the node performance monitoring value, then calculates the second CPU occupancy ratio according to the second CPU occupied resource amount and the preset CPU occupied resource amount, then calculates the second RAM occupancy ratio according to the second RAM occupied resource amount and the preset RAM occupied resource amount, and at the same time calculates the second GPU occupancy ratio according to the second GPU occupied resource amount and the preset GPU occupied resource amount, then calculates the second bandwidth occupancy ratio according to the second bandwidth occupied resource amount and the preset bandwidth occupied resource amount, and finally, calculates the node resource occupancy score according to the second CPU occupancy ratio, the second RAM occupancy ratio, the second GPU occupancy ratio and the second bandwidth occupancy ratio, so that the system can optimize resources and make decisions through the node resource occupancy score, thereby providing an important basis for subsequent resource scheduling.

[0030] In one embodiment, the step S6 of dynamically scheduling the computing and storage network resources required by the corresponding node according to the first resource scheduling instruction further includes: S601, traversing and matching the resource occupancy score of the application task with the resource occupancy scores of multiple nodes based on a similarity model to obtain a resource similarity value, wherein the function of the similarity model is: ; in, Indicates the resource similarity value, Indicates the resource usage score of the application task. It represents the node resource usage score, n represents the sequence number of the node resource usage score, n=1, 2, 3...n.

[0031] As described in the above step S601, the present invention traverses and matches the application task resource occupancy score with multiple node resource occupancy scores based on a similarity model to obtain a resource similarity value. When the resource similarity value is closer to 1, it means that the two are more matched. In this way, the system can quickly match task resources and node resources, and make the best resource allocation decision based on real-time data. This method can ensure that resources are used efficiently, and at the same time, it can also ensure that tasks are executed under appropriate conditions, thereby improving the overall performance and response speed of the system, and is suitable for scheduling long-term edge tasks.

[0032] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps for the OpenSAI intelligent network system and the computing network matching method are implemented.

[0033] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0034] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0035] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An OpenSAI intelligent network system, characterized in that: include: Application task module, cluster management module, computing distribution module, adaptation optimization module, resource monitoring module and resource control module; The application task module is used to obtain multiple application tasks, sort the multiple application tasks to obtain a task sorting list, and perform interface parameter resource configuration on the multiple application tasks in sequence according to the task sorting list to obtain multiple application task configuration resources, and send the multiple application task configuration resources to the computing cluster management module; The cluster management module is used to obtain the current node status and configure resources to form a computing node cluster according to the current node status and the multiple application tasks; The computing distributed module is used to offload multiple application tasks to the network nodes in the computing node cluster for execution, thereby obtaining multiple execution tasks; The adaptation optimization module is used to dynamically configure the plurality of execution tasks based on the resources provided by the elastic base, wherein the dynamic configuration includes application task configuration, application task and network node configuration and network resource scheduling configuration; The resource monitoring module is used to monitor the status of the network nodes in the computing node cluster that have been offloaded with multiple application tasks, obtain multiple node performance monitoring values, and pass the multiple node performance monitoring values ​​to the adaptation optimization module, and determine in turn whether the multiple node performance monitoring values ​​meet the preset monitoring values; If not, generating a first resource scheduling instruction of the corresponding required resource for the node that does not meet the preset monitoring value; The resource control module is used to receive a first resource scheduling instruction and dynamically schedule the computing and storage network resources required by the corresponding node according to the first resource scheduling instruction.

2. The OpenSAI intelligent network system according to claim 1, characterized in that: The adaptation optimization module is also used to simultaneously optimize the application task configuration and network bandwidth and network computing resources, obtain task-network bidirectional synchronization optimization information, and pair the application tasks and network nodes based on the task-network bidirectional synchronization optimization information.

3. A computing network matching method is used to execute the OpenSAI intelligent network system according to any one of claims 1-2, characterized in that: include: Acquire multiple application tasks, sort the multiple application tasks to obtain a task sorting list, and sequentially configure interface parameter resources for the multiple application tasks according to the task sorting list to obtain multiple application task configuration resources; Obtaining the current node status, and configuring resources to form a computing node cluster according to the current node status and the plurality of application tasks; Offloading multiple application tasks to network nodes in a computing node cluster for execution, thereby obtaining multiple execution tasks; Dynamically configuring the plurality of execution tasks based on resources provided by the elastic base, wherein the dynamic configuration includes application task configuration, application task and network node configuration, and network resource scheduling configuration; Monitor the status of the network nodes in the computing node cluster to which the multiple application tasks are offloaded, obtain multiple node performance monitoring values, and determine in turn whether the multiple node performance monitoring values ​​meet the preset monitoring values; If not, generating a first resource scheduling instruction of the corresponding required resource for the node that does not meet the preset monitoring value; The computing and storage network resources required by the corresponding node are dynamically scheduled according to the first resource scheduling instruction.

4. The network matching method according to claim 3, characterized in that: The step of sequentially configuring interface parameter resources for the plurality of application tasks according to the task sorting list to obtain a plurality of application task configuration resources also includes: Acquire a first CPU occupied resource amount, a first RAM occupied resource amount, a first GPU occupied resource amount, and a first bandwidth occupied resource amount of the application task according to the application task configuration resources; Calculating a first CPU occupancy ratio according to the first CPU occupied resource amount and a preset CPU occupied resource amount; Calculating a first RAM occupancy ratio according to the first RAM occupied resource amount and a preset RAM occupied resource amount; Calculating a first GPU occupancy ratio according to the first GPU occupied resource amount and the preset GPU occupied resource amount; Calculating a first bandwidth occupancy ratio according to the first bandwidth occupied resource amount and a preset bandwidth occupied resource amount; The resource occupancy score of the application task is calculated according to the first CPU occupancy ratio, the first RAM occupancy ratio, the first GPU occupancy ratio, and the first bandwidth occupancy ratio, wherein the calculation formula is: ； in, Indicates the resource usage score of the application task. Indicates the first CPU usage ratio. Indicates the first RAM occupancy ratio, Indicates the first GPU occupancy rate, Indicates the first bandwidth occupancy ratio; According to the above steps, multiple application task resource occupancy scores corresponding to the multiple application task configuration resources are calculated in sequence.

5. The network matching method according to claim 3, characterized in that: The step of monitoring the status of the network nodes in the computing node cluster to obtain multiple node performance monitoring values ​​according to the multiple application tasks being offloaded to the computing node cluster further includes: Obtaining a second CPU occupied resource amount, a second RAM occupied resource amount, a second GPU occupied resource amount, and a second bandwidth occupied resource amount of the node according to the node performance monitoring value; Calculate the second CPU occupancy ratio according to the second CPU occupied resource amount and the preset CPU occupied resource amount; Calculate a second RAM occupancy ratio according to the second RAM occupied resource amount and a preset RAM occupied resource amount; Calculating the second GPU occupation ratio according to the second GPU occupation amount and the preset GPU occupation amount; Calculating a second bandwidth occupation ratio according to the second bandwidth occupied resource amount and a preset bandwidth occupied resource amount; The node resource occupancy score is calculated according to the second CPU occupancy ratio, the second RAM occupancy ratio, and the second bandwidth occupancy ratio, wherein the calculation formula is: ； in, Indicates the node resource usage score, Indicates the second CPU usage ratio. Indicates the second RAM occupancy ratio, Indicates the first GPU occupancy rate, represents the second bandwidth occupancy ratio; According to the above calculation steps, multiple node resource occupancy scores corresponding to the multiple node performance monitoring values ​​are calculated in sequence.

6. The network matching method according to claim 3, characterized in that: The step of dynamically scheduling the computing and storage network resources required by the corresponding node according to the first resource scheduling instruction also includes: The resource occupancy score of the application task is traversed and matched with the resource occupancy scores of multiple nodes based on the similarity model to obtain a resource similarity value, wherein the function of the similarity model is: ; in, Indicates the resource similarity value, Indicates the resource usage score of the application task. It represents the node resource usage score, n represents the sequence number of the node resource usage score, n=1, 2, 3...n.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 3 to 6 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 3 to 6 are implemented.

Images