Computer and cloud computing power optimization method based on multi-task collaboration

Abstract

The invention relates to the technical field of computing power optimization, and provides a computer and cloud computing power optimization method based on multi-task collaboration. The method comprises the following steps: receiving Q to-be-executed tasks of a local cloud collaborative network; performing priority constraint on the Q to-be-executed tasks according to the multivariate priority evaluation factors, and establishing a task chain; collecting callable computing resource information of a local cloud collaborative network, and establishing a resource space; based on the resource space, performing task chain resource allocation by utilizing a computing resource scheduling optimization algorithm to obtain a scheduling strategy chain; performing cloud computing power balance optimization according to the scheduling strategy chain to obtain an optimized chain; and executing the task chain according to the optimization chain based on the cloud collaborative network. The technical problems of unbalanced scheduling and low resource utilization rate between the cloud and the local computing resources are solved, and the technical effects of improving the cooperative work efficiency and the resource utilization rate of the cloud and the local computing resources through priority constraint and computing resource scheduling optimization are achieved.

Description

Technical Field

[0001] The present application relates to the technical field of computing power optimization, and specifically to a method for optimizing computer and cloud computing power based on multi-task collaboration. Background Art

[0002] With the rapid development of information technology, the hybrid computing model combining cloud computing and local computing has been increasingly widely used in large-scale data processing, artificial intelligence applications, and high-performance computing. In this context, local cloud collaborative networks, as an effective way to expand computing power, have become a key technology to promote the optimal configuration of computing resources. However, when dealing with multi-task collaboration, the current computing resource scheduling system has problems such as uneven distribution of computing resources, inaccurate task priority management, and low computing power utilization. Existing cloud computing scheduling methods often focus on the scheduling of a single computing platform or task, but lack comprehensive consideration of the collaborative work of local computers and cloud computing platforms. In addition, due to the diversity of tasks and the limited resources, traditional resource scheduling methods are difficult to cope with complex computing requirements when executing multiple tasks in a collaborative manner, resulting in waste of computing power and increased latency. Summary of the invention

[0003] This application aims to solve the technical problems of unbalanced scheduling and low resource utilization between cloud and local computing resources by providing a computer and cloud computing power optimization method based on multi-task collaboration.

[0004] In view of the above problems, the present application provides a computer and cloud computing power optimization method based on multi-task collaboration.

[0005] The present application provides a computer and cloud computing power optimization method based on multi-task collaboration, the method is applied to a local cloud collaborative network, the local cloud collaborative network includes multiple local computers and a cloud computing platform, the method includes: receiving Q tasks to be executed of the local cloud collaborative network, wherein Q is a positive integer greater than 1; performing priority constraints on the Q tasks to be executed according to a multi-priority evaluation factor, and establishing a task chain to be executed; collecting the callable computing resource information of the local cloud collaborative network, and establishing a local cloud computing resource space; based on the local cloud computing resource space, allocating computing resources to the task chain to be executed according to a computing resource scheduling optimization algorithm, and obtaining a computing resource scheduling strategy chain; performing computing power balancing optimization of the local cloud collaborative network according to the computing resource scheduling strategy chain, and obtaining a computing resource scheduling optimization chain; based on the local cloud collaborative network, executing the task chain to be executed according to the computing resource scheduling optimization chain.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: The above-mentioned computer and cloud computing power optimization method based on multi-task collaboration aims to improve computing resource utilization and task execution efficiency by efficiently scheduling and optimizing resources for multiple tasks to be executed in the local cloud collaborative network; first, Q tasks to be executed are received, where Q is greater than 1, indicating that multiple tasks need to be processed. In this process, tasks are prioritized by multiple priority evaluation factors to ensure that high-priority tasks can be executed first, thereby effectively avoiding task delays and resource conflicts; then, the callable computing resource information in the local cloud collaborative network is collected to construct a local cloud computing resource space. In this way, the distribution and availability of computing resources are accurately understood, providing comprehensive data support for subsequent scheduling decisions. ; Then, based on this computing resource space, the computing resource scheduling optimization algorithm is used to optimize the allocation of the task chain to be executed, ensuring that each task can obtain the most appropriate computing resource support according to its needs, reducing task waiting or execution delays caused by insufficient resources, thereby improving task processing efficiency; with the generation of the computing resource scheduling strategy chain, computing power balance optimization is performed according to these strategies to ensure that the allocation of computing resources is more balanced and effective; then, by adjusting the scheduling optimization chain of computing resources, computing power overload or idleness is avoided, and efficient use of computing resources is guaranteed; finally, based on the optimized computing resource scheduling strategy chain, the task chain to be executed is executed to ensure that each task obtains appropriate resource support at the most appropriate time, achieving the best balance between task execution and resource scheduling.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 It is a flowchart of a method for optimizing computer and cloud computing power based on multi-task collaboration in one embodiment; Figure 2 The present invention is a flowchart of a computing resource scheduling optimization algorithm of a computer and cloud computing power optimization method based on multi-task collaboration in one embodiment. DETAILED DESCRIPTION

[0010] The embodiments of the present application solve the technical problems of unbalanced scheduling and low resource utilization between cloud and local computing resources by providing a computer and cloud computing power optimization method based on multi-task collaboration.

[0011] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0012] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0013] Examples, such as Figure 1 As shown, the present application provides a computer and cloud computing power optimization method based on multi-task collaboration, the method is applied to a local cloud collaboration network, the local cloud collaboration network includes multiple local computers and a cloud computing platform, the method includes: Receive Q tasks to be executed from the local cloud collaborative network, where Q is a positive integer greater than 1.

[0014] The method is applied to a local cloud collaborative network, which is composed of multiple local computers and a cloud computing platform. The local computers are responsible for processing local data and tasks, while the cloud computing platform provides more powerful computing resources to process larger-scale and more complex tasks. By combining local computing resources with cloud computing resources to form a collaborative network, the most suitable computing platform can be dynamically selected to perform tasks under different tasks and computing requirements, thereby improving overall computing efficiency. In this network architecture, local computers can handle some tasks that are more urgent or require low latency, while the cloud computing platform can undertake tasks that require larger computing resources or have longer computing cycles. The collaborative work of the two enables computing resources to be more evenly distributed, while avoiding the situation where the computing resources of one party are overloaded or idle.

[0015] In an embodiment of the present application, Q tasks to be executed are received from a local cloud collaborative network, where Q represents the number of tasks and is greater than 1. These tasks may come from different sources and involve different computing requirements and execution priorities. Therefore, effective management and scheduling are required among multiple tasks to be executed. The purpose of receiving these tasks is to understand the total number of tasks to be processed and to lay the foundation for subsequent task priority division, resource scheduling, and computing power allocation.

[0016] Priority constraints are imposed on the Q tasks to be executed according to the multi-priority evaluation factors, and a task chain to be executed is established.

[0017] In one embodiment, the Q tasks to be executed are prioritized and constrained according to the multi-priority evaluation factors to ensure that important or urgent tasks can be processed first, wherein the multi-priority evaluation factors may include the urgency of the task, the importance of the task, and the delay loss degree, etc. These factors work together to determine the execution priority of each task; based on these priority evaluation factors, the Q tasks to be executed are sorted according to the priority, thereby establishing a task chain to be executed, which is an ordered task queue, in which each task has a clear execution order; the task chain to be executed includes the first task to be executed, the second task to be executed... until the Qth task to be executed, and the sorted task chain can reflect the priority relationship between tasks, ensuring that in the case of limited resources, high-priority tasks are executed first to avoid delays in low-priority tasks. Through this reasonable sorting method based on the actual needs and priorities of the tasks, each task can be executed at the most suitable time, and this priority constraint helps to optimize the overall task execution process, improve the efficiency of resource use and the execution speed of tasks.

[0018] Furthermore, the present application provides a method of performing priority constraints on the Q tasks to be executed according to the multi-priority evaluation factors, and establishing a task chain to be executed, including: The multi-priority evaluation factors include task urgency, task importance and task delay loss; the Q tasks to be executed are prioritized according to the multi-priority evaluation factors to obtain Q task execution priority coefficients; the Q tasks to be executed are prioritized according to the Q task execution priority coefficients to establish an initial task chain; the dependency information of the Q tasks to be executed is collected to construct a task dependency graph; according to the task dependency graph, the initial task chain is optimized to generate the task chain to be executed.

[0019] Preferably, the Q tasks to be executed are evaluated according to multiple priority evaluation factors, and these evaluation factors include task urgency, task importance and task delay loss, wherein the task urgency is used to determine whether the task needs to be processed immediately, and tasks with high urgency are executed first, the task importance is the importance determined according to the degree of impact of the task on the overall business objectives, and tasks with high importance will be given higher priority, and the task delay loss is used to evaluate the possible loss if the task is delayed, and tasks with higher delay loss will be executed first to avoid risks; then, according to these evaluation factors, the Q tasks to be executed are prioritized through the task priority evaluation weighted network, and the execution priority coefficient of each task to be executed is calculated, and these coefficients represent the execution priority of the tasks; then, the priority coefficients of all the tasks to be executed are compared, and the tasks to be executed corresponding to each task execution priority coefficient are sorted in descending order, and the sorted task chain is the initial task chain, and the tasks on this chain have been sorted according to their priority relationship, with urgent and important tasks in front and lower priority tasks in the back. ; Then, the dependency information of each task is collected. This step is to identify which tasks can only be started after other tasks are completed. For example, task A may depend on the result of task B. Task A can only be started after task B is completed. Dependency information is a key factor for mutual constraints between tasks; Based on the collected task dependency information, a task dependency graph is constructed. Each node in the graph represents a task, and the edge represents the dependency between tasks. The structure of the graph clearly shows the sequence and dependency constraints between tasks; Based on the task dependency graph, the initial task chain is optimized, that is, through the dependency between tasks, the tasks to be executed in the initial task chain that are contrary to the dependency are adjusted in order to ensure that the execution order of the tasks meets the dependency constraints and at the same time, the execution efficiency of the tasks is improved as much as possible. For example, the task dependency graph can help avoid deadlock in task execution and ensure the smooth progress of tasks; After dependency optimization, the generated task chain to be executed takes into account the priority of the tasks and ensures that the dependencies between tasks are satisfied, and finally forms a task execution plan that is both efficient and in line with the logical order. Through the above steps, not only can the execution order be reasonably arranged according to the priority of the tasks, but also the dependencies between tasks can be ensured to be complied with, thereby optimizing the entire task scheduling process and improving the utilization efficiency of computing resources and the execution efficiency of tasks.

[0020] Furthermore, the present application provides a method of performing priority evaluation on the Q tasks to be executed according to the multi-priority evaluation factor to obtain Q task execution priority coefficients, including: The Q tasks to be executed are prioritized according to the multi-priority evaluation factors to obtain Q task priority evaluation results, wherein each task priority evaluation result includes a task urgency coefficient, a task importance coefficient and a task delay loss coefficient corresponding to each task to be executed; weights are allocated according to the multi-priority evaluation factors to establish a task priority evaluation weighted network, wherein the task priority evaluation weighted network includes a task urgency weight, a task importance weight and a task delay loss weight; the Q task priority evaluation results are input into the task priority evaluation weighted network to obtain the Q task execution priority coefficients.

[0021] Optionally, in order to evaluate the priority of the tasks to be executed, a priority evaluation model will be pre-built. This model includes a task urgency evaluation channel, a task importance evaluation channel, and a task delay loss evaluation channel. The task urgency evaluation channel is a mathematical model with a built-in urgency coefficient calculation formula. The formula calculates the task urgency coefficient by calculating the ratio of the remaining time of the task to the total task duration. The task importance evaluation channel and the task delay loss evaluation channel are machine learning models, which are trained through their own historical data in the same way. Taking the task importance evaluation channel as an example, historical task execution data is collected from historical task execution logs, including task type (core task, supporting task, auxiliary task, etc.), task dependency (is it The task execution data are divided into training set and test set according to the task characteristics (whether it is a critical path task), the task stage (demand analysis, development, testing, delivery, etc.), and the importance coefficients marked by professional technicians in the industry. Then, the multi-layer perceptron (MLP) is selected as the task importance evaluation channel. Its input layer receives various features of the task, and multiple hidden layers are responsible for feature extraction and nonlinear mapping. The output layer generates the task importance coefficient. After that, it enters the training phase. In this phase, the task feature data in the training set (except the data with marked importance coefficients) are used to train the task importance evaluation channel. The training process includes forward propagation, loss calculation, back propagation and parameter optimization. In the forward propagation phase, the task feature data is input into the test set. The task importance evaluation channel is input, and the final output is obtained through weighted operations and activation function calculations of each layer; in the loss calculation stage, the mean square error (MSE) is used as the loss function to measure the gap between the task importance coefficient predicted by the task importance evaluation channel and the true value, and then the back propagation algorithm is used to calculate the gradient of the loss function for each parameter, and the parameters are updated through the Adam optimizer to continuously adjust the weights and biases of the task important evaluation channel until the training process converges; after the training is completed, the model is evaluated. In this stage, the data in the test set is used to evaluate the performance of the trained task important evaluation channel, the error between the predicted value and the true value is calculated, and the prediction accuracy of the model is measured through regression evaluation indicators such as the mean square error; if The model performs well on the test set (i.e. the error is less than the preset error), indicating that the task importance evaluation channel has understood the relationship between task characteristics and importance coefficients and has strong generalization ability; after constructing the task urgency evaluation channel, task importance evaluation channel, and task delay loss evaluation channel, these three channels are connected in parallel to construct a priority evaluation model; by inputting the task feature data of Q tasks to be executed into the priority evaluation model, the task urgency coefficient, task importance coefficient, and task delay loss coefficient of each task to be executed are calculated through the task urgency evaluation channel, task importance evaluation channel, and task delay loss evaluation channel within the model, and these coefficients are stored accordingly to form Q task priority evaluation results;After completing the priority evaluation of the task, weights are assigned to these factors according to the relative importance of each factor. The weight allocation is based on historical experience and expert advice, reflecting the degree of influence of each evaluation factor in the priority sorting. Through the reasonable allocation of these three weights, a task priority evaluation weighted network is established. This network combines the priority factors (urgency, importance, delay loss) of each task according to the weights, so as to obtain a more accurate priority sorting; after completing the weight allocation and establishing the task priority evaluation weighted network, the priority evaluation results of each task are input into the weighted network. The task priority evaluation weighted network performs weighted calculations on the three priority factors (urgency, importance, delay loss) of each task through the task urgency weight, task importance weight and task delay loss weight to obtain the task execution priority coefficient of each task. This coefficient reflects the priority of the task to be executed among all the tasks to be executed. The higher the value, the higher the task should be executed first. ;

[0022] Collect the available computing resource information of the local cloud collaborative network and establish a local cloud computing resource space.

[0023] In one embodiment, first, information on available computing resources in a local cloud collaborative network is collected and obtained. These computing resources include but are not limited to local computers, cloud servers, storage devices, network bandwidth, etc., all resources that can be scheduled and utilized during task execution. Specifically, the available computing resource information may include computing nodes, storage resources, bandwidth resources, resource status, etc., wherein computing nodes refer to device resource configurations possessed by local computers or cloud servers, such as CPU, GPU, memory, and other computing resources; storage resources refer to cloud or local storage devices, including disk storage, cloud storage, etc., and their current usage, such as remaining space; bandwidth resources refer to the availability of network bandwidth, including bandwidth resources between the local network and the cloud and local networks, which affect data transmission speed and communication delay during task execution; resource status refers to the current load status of each computing node or resource pool, which resources are being used, which resources are idle, and when resources are available. Long, etc.; by collecting this data, we can understand the status and capabilities of each computing resource in the current local cloud collaborative network, and provide a basis for subsequent task scheduling; after obtaining the information of these callable computing resources, integrate and organize them to establish a local cloud computing resource space. This computing resource space is a collection that comprehensively displays all available resources, including local computing resources and cloud computing resources. Local computing resources refer to the computing nodes, storage resources, bandwidth resources, and resource status provided by local computers and internal servers. Cloud computing resources refer to the computing nodes, storage resources, bandwidth resources, and resource status provided by cloud service providers; by establishing a local cloud computing resource space, these distributed resources can be managed in a unified manner, and the use of resources can be obtained in real time, so as to make optimization decisions during task scheduling and resource allocation, so as to flexibly schedule resources between local and cloud, achieve more efficient computing resource utilization, and improve the performance and response speed of task execution.

[0024] Based on the local cloud computing resource space, computing resources are allocated to the task chain to be executed according to a computing resource scheduling optimization algorithm to obtain a computing resource scheduling strategy chain.

[0025] In one embodiment, the constructed local cloud computing resource space is used to allocate the required computing resources for the task chain to be executed. The computing resource scheduling process is based on a computing resource scheduling optimization algorithm. The core goal of the algorithm is to optimize resource scheduling through a computing resource scheduling optimization module according to the specific requirements of each task to be executed (such as computing node requirements, storage requirements, bandwidth requirements, etc.) and the available resources in the local cloud computing resource space (such as computing nodes, storage resources, bandwidth resources, etc.), and then update the local cloud computing resource space according to the optimization results. Based on the updated local cloud computing resource space, the same operation is performed on the subsequent second to Qth tasks to be executed to obtain the computing resource scheduling strategy for each task to be executed; finally, all computing resource scheduling strategies are summarized to generate a set of computing resource scheduling strategy chains, which lists in detail the computing resources that should be allocated to each task to be executed, which helps to ensure the effective use of resources, reduce resource competition in task execution, and improve the overall efficiency and response speed of task execution.

[0026] Further, if Figure 2 As shown, the present application provides the computing resource scheduling optimization algorithm, including: Input the first task to be executed in the chain of tasks to be executed into the computing resource demand analysis model to obtain the computing resources required by the first task; take the computing resources required by the first task as the resource scheduling target, make a resource scheduling decision on the local cloud computing resource space, and obtain a first task resource scheduling decision domain; based on the first task to be executed, optimize the first task resource scheduling decision domain according to the computing resource scheduling optimization module to obtain a first computing resource scheduling strategy; update the local cloud computing resource space according to the first computing resource scheduling strategy, and continue to optimize the computing resource scheduling of the second task to be executed...the Qth task to be executed in the chain of tasks to be executed in combination with the computing resource demand analysis model and the computing resource scheduling optimization module to obtain a second computing resource scheduling strategy...the Qth computing resource scheduling strategy; construct the computing resource scheduling strategy chain according to the first computing resource scheduling strategy, the second computing resource scheduling strategy...the Qth computing resource scheduling strategy.

[0027] Preferably, the first task in the task chain to be executed is input into a pre-built computing resource demand analysis model, with the purpose of obtaining the computing resources required for the task, including the required computing nodes, required storage resources, required bandwidth resources, etc. The model is based on historical task feature data (such as task type, task equipment requirements, task dependencies, task estimated completion time, task output type, etc.) and historical task resource input data (such as involved computing nodes, used storage resources, used bandwidth resources) collected from historical task execution logs, and is trained through the same forward propagation, loss calculation, back propagation, parameter optimization and other steps as mentioned above, and can evaluate the amount of resources required for the task according to the characteristics of the task; then , the computing resources required by the first task generated by the computing resource demand analysis model are used as the target of resource scheduling. This target is to minimize the difference between task requirements and available resources, that is, to ensure that tasks can get reasonable resource allocation under limited resources; according to this resource scheduling target, the scheduling decision of computing resources is started, that is, the available resources in the local cloud computing resource space are randomly combined to meet the requirements of the resource scheduling target. For example, assuming that a task to be executed requires 4 CPU cores, 8GB of memory and 50GB of storage, 3 CPU cores, 6GB of memory and 30GB of storage can be allocated from the local, and 1 CPU core, 2GB of memory and 20GB of storage can be allocated from the cloud, or all of them can be allocated from the cloud. The resource allocation is 5 CPU cores, 10GB memory, and 60GB storage (although the resources are in excess, the priority of the task can be guaranteed). It can also balance local and cloud resources, that is, 2 CPU cores, 4GB memory, and 25GB storage come from the local, and 2 CPU cores, 4GB memory, and 25GB storage come from the cloud; after obtaining the first task resource scheduling decision domain, the computing resource scheduling optimization module will further optimize the resource scheduling based on this first task resource scheduling decision domain. The module performs expected optimization and iterative optimization through the internal resource scheduling prediction unit and resource scheduling compatibility calculation unit to select the optimal task resource scheduling decision to ensure that the task maximizes the resource while meeting its computing resource requirements. In order to improve utilization efficiency and reduce idleness and waste of resources, the optimized first computing resource scheduling strategy will specify the specific computing resources allocated to the first task; then, according to the obtained first computing resource scheduling strategy, the local cloud computing resource space is updated, that is, after the resources are allocated, the remaining available amount of each resource will be recalculated (for example, the CPU, memory, storage, etc. allocated to the task will be subtracted from the original remaining available amount). This update process is dynamic. After each allocation of resources for a task, the computing resource space must be updated to prepare for the allocation of the next task; after updating the resource space, the remaining tasks to be executed (such as the second task to be executed, the third task to be executed, and the Qth task to be executed) will be scheduled for computing resources in sequence;For each subsequent task, the same process is still used, that is, the computing resources required by the task are obtained through the computing resource demand analysis model, and resource scheduling is performed in the updated computing resource space. After optimization, the computing resource scheduling strategy of each task to be executed is obtained; finally, the computing resource scheduling strategies of all tasks to be executed will be arranged based on the order of each task to be executed in the task chain to be executed, and a complete computing resource scheduling strategy chain will be constructed. This strategy chain lists the computing resource scheduling strategies of all tasks in detail, providing a clear resource scheduling basis for the execution of subsequent tasks to be executed, ensuring that each task to be executed can obtain reasonable computing resources, and the resource utilization during the execution of the task to be executed is maximized to avoid resource conflicts and waste. ;

[0028] Further, the present application provides, based on the first task to be executed, optimizing the first task resource scheduling decision domain according to the computing resource scheduling optimization module to obtain a first computing resource scheduling strategy, including: The computing resource scheduling optimization module includes a resource scheduling prediction unit and a resource scheduling compatibility calculation unit; based on the first task to be executed, a first task resource scheduling expectation is set, wherein the first task resource scheduling expectation includes the expected resource utilization, expected task execution efficiency and expected task execution quality coefficient corresponding to the first task to be executed; based on the resource scheduling prediction unit, the first task resource scheduling decision domain is optimized according to the first task resource scheduling expectation to obtain the first task resource scheduling optimization domain; according to the resource scheduling compatibility calculation unit, the first task resource scheduling optimization domain is optimized iteratively to generate the first computing resource scheduling strategy.

[0029] Optionally, the core of optimizing the resource scheduling decision domain of the first task is to optimize the allocation of resources according to the expected scheduling requirements of the task through the resource scheduling prediction unit and the resource scheduling compatibility calculation unit of the computing resource scheduling optimization module, wherein the resource scheduling prediction unit is divided into a resource utilization prediction model, a task execution efficiency prediction model and a task execution quality prediction model. These three prediction models are all constructed based on the historical task feature data, historical task resource input data and corresponding identifiers (resource utilization, task execution efficiency, task execution quality coefficient) extracted from the historical task execution logs through the same training method as mentioned above. The resource scheduling compatibility calculation unit is a mathematical model. The pre-configured weights are used to perform a weighted summation of the input resource utilization, task execution efficiency, and task execution quality coefficient to quantify the suitability of the task resource scheduling decision. When optimizing the first task resource scheduling decision domain, a set of expected resource scheduling indicators is first set for the first task to be executed. These expected values ​​provide guidance for task scheduling to ensure that the scheduling results meet the efficiency and quality requirements of task execution. The set expectations include expected resource utilization, expected task execution efficiency, and expected task execution quality coefficient. Among them, the expected resource utilization is the expected efficiency of the resource (such as CPU, memory, storage, etc.) used by the task to be executed, and the expected task execution efficiency is the expected execution speed of the task to be executed after the resources are allocated. The expected task execution quality coefficient is the execution quality after the task to be executed is completed. These expected values ​​are determined based on task requirements, historical experience and industry standards, and provide targets for subsequent scheduling and optimization. After that, based on the resource scheduling prediction unit, each resource scheduling decision in the first task resource scheduling decision domain is evaluated to obtain the predicted resource utilization rate, predicted task execution efficiency and predicted task execution quality coefficient of each resource scheduling decision. Then, these prediction results are analyzed in correspondence with the first task resource scheduling expectations, and the resource scheduling decisions that meet the first task resource scheduling expectations are selected and added to the first task resource scheduling optimization domain. After obtaining the first task resource scheduling optimization domain, The optimization domain is further optimized through the resource scheduling compatibility calculation unit. In this process, the resource scheduling compatibility calculation unit will evaluate the compatibility of each resource scheduling decision in the first task resource scheduling optimization domain. The compatibility calculation involves the comprehensive calculation of indicators such as resource utilization, execution efficiency and quality. Based on the compatibility calculation results, each plan in the resource scheduling decision domain is iteratively optimized to improve the matching degree. Through multiple rounds of iterative optimization, it gradually converges to the optimal plan, thereby obtaining the first computing resource scheduling strategy, that is, an optimal resource scheduling plan, which can meet the expected resource utilization, execution efficiency and quality requirements of the task to be executed, while avoiding excessive waste of resources to ensure efficient execution of tasks.

[0030] Further, the present application provides a method of performing expected optimization on the first task resource scheduling decision domain based on the resource scheduling prediction unit according to the first task resource scheduling expectation to obtain the first task resource scheduling optimization domain, including: The resource scheduling prediction unit includes a resource utilization prediction model, a task execution efficiency prediction model and a task execution quality prediction model; according to the first task resource scheduling decision domain, a first resource scheduling decision is extracted; the first resource scheduling decision and the first task to be executed are input into the resource scheduling prediction unit to obtain a first resource scheduling prediction result; it is determined whether the first resource scheduling prediction result meets the first task resource scheduling expectation; if the first resource scheduling prediction result meets the first task resource scheduling expectation, the first resource scheduling decision is added to the first task resource scheduling optimization domain; according to the resource scheduling prediction unit and the first task resource scheduling expectation, the first task resource scheduling decision domain is continued to be optimized, and the first task resource scheduling optimization domain is constructed.

[0031] Optionally, the resource scheduling prediction unit includes a resource utilization prediction model, a task execution efficiency prediction model and a task execution quality prediction model, which work together to provide prediction support for resource scheduling, wherein the resource utilization prediction model is used to predict the degree of utilization of computing resources by the task to be executed under a given resource allocation, and helps determine whether the resources allocated to the task can be efficiently utilized to avoid resource waste or resource shortage; the task execution efficiency prediction model is used to predict the execution efficiency or execution time of the task to be executed under a given resource configuration, and helps evaluate whether the task to be executed can be completed within the scheduled time, or whether the task to be executed can be completed within a certain resource condition. The task execution quality prediction model is used to predict the execution quality of the task to be executed under a given resource configuration, to help evaluate whether the task to be executed can meet the expected execution standard, and to ensure that the quality of the task to be executed will not be reduced due to unreasonable resource allocation when it is completed; in the process of expectation optimization, firstly, according to the generated first task resource scheduling decision domain, a resource scheduling decision is randomly extracted as the first resource scheduling decision. The resource scheduling decision domain is a set containing multiple candidate resource scheduling decisions, each of which meets the resource scheduling goal of the task, but they are different in resource configuration, load distribution, etc.; then, the selected first resource The resource scheduling decision and the characteristic data of the first task to be executed are input into the resource scheduling prediction unit. The resource scheduling prediction unit predicts the utilization rate of the task resources through the internal resource utilization prediction model, predicts the execution efficiency of the task through the task execution efficiency prediction model, and predicts the execution quality of the task through the task execution quality prediction model. Then, the resource utilization prediction value, task execution efficiency prediction value and task execution quality coefficient prediction value of the first task to be executed output by the three prediction models are added to the first resource scheduling prediction result and output. After obtaining the prediction result, it is compared with the first task resource scheduling expectation of the first task to be executed. If the resource utilization prediction value in the prediction result is higher than the expected value, the prediction value of the task execution efficiency and the prediction value of the task execution quality coefficient are higher than the expected value, the prediction value of the task execution quality coefficient are higher than the expected value, the prediction value of the task execution quality coefficient are higher than the expected value, and the prediction value of the task execution quality coefficient is higher than the expected value. If the resource utilization, task execution efficiency and execution quality meet the expected values, the resource scheduling decision is considered to be appropriate. At this time, this resource scheduling decision is added to the first task resource scheduling optimization domain and the next resource scheduling decision is evaluated. Otherwise, this resource scheduling decision is discarded and the next resource scheduling decision is evaluated. Repeat the above process and continue to perform expected optimization on the first task resource scheduling decision domain through the resource scheduling prediction unit and the first task resource scheduling expectation, so as to construct the final first task resource scheduling optimization domain, ensuring that the task to be executed can be executed in an efficient and high-quality manner under given resource conditions, while avoiding resource waste and scheduling delays.

[0032] Furthermore, the present application provides, according to the resource scheduling compatibility calculation unit, performing optimization iteration on the first task resource scheduling optimization domain to generate the first computing resource scheduling strategy, including: According to the first task resource scheduling optimization domain, extract the nth resource scheduling decision, where n is a positive integer; input the nth resource scheduling prediction result corresponding to the nth resource scheduling decision into the resource scheduling compatibility calculation unit to obtain the nth resource scheduling compatibility coefficient, where the resource scheduling compatibility calculation unit includes resource utilization weight, task execution efficiency weight and task execution quality weight; according to the first task resource scheduling optimization domain, extract the n+1th resource scheduling decision, and calculate the n+1th resource scheduling compatibility coefficient in combination with the resource scheduling compatibility calculation unit; according to the nth resource scheduling compatibility coefficient and the n+1th resource scheduling compatibility coefficient, perform current scheduling compatibility optimization on the nth resource scheduling decision and the n+1th resource scheduling decision to obtain the current optimized resource scheduling decision; based on the resource scheduling compatibility calculation unit and the first task resource scheduling optimization domain, iteratively optimize the current optimized resource scheduling decision until the number of iterative optimizations meets the iterative optimization number constraint, and obtain the first computing resource scheduling strategy.

[0033] Optionally, extract the nth (an integer) resource scheduling decision from the first task resource scheduling optimization domain, and then transmit the nth resource scheduling prediction result corresponding to the nth resource scheduling decision to the resource scheduling prediction unit, and calculate the nth resource scheduling fit coefficient. This resource scheduling fit coefficient is obtained by weighted calculation of the resource utilization prediction value, the task execution efficiency prediction value and the task execution quality coefficient prediction value through the built-in resource utilization weight, the task execution efficiency weight and the task execution quality weight, and represents the degree of match between the nth resource scheduling decision and the expected effect in meeting the task resource requirements. Subsequently, extract the n+1th resource scheduling decision from the first task resource scheduling optimization domain, and calculate the n+1th resource scheduling fit coefficient in the same way, and then compare the nth resource scheduling fit coefficient with the n+1th resource scheduling fit coefficient. Perform current scheduling fit optimization, that is, select the resource scheduling decision corresponding to the largest resource scheduling fit coefficient as the current optimal resource scheduling decision. This decision can optimize resource utilization and task execution efficiency to the greatest extent while meeting the task resource requirements; then, through the resource scheduling fit calculation unit, continue to iteratively optimize the current optimal resource scheduling decision. Each iteration is based on the resource scheduling decision and fit coefficient of the previous round, and further adjustments are made to better meet the expected requirements of the task. The iterative process continues until the predetermined number of iterative optimization constraints is reached; finally, through multiple rounds of optimization iterations, the first computing resource scheduling strategy is generated. This strategy has the largest resource scheduling fit coefficient, which ensures that the task to be executed can be completed within a reasonable time, while maximizing the utilization efficiency of resources and ensuring the quality of task execution.

[0034] The computing power balance optimization of the local cloud collaborative network is performed according to the computing resource scheduling strategy chain to obtain a computing resource scheduling optimization chain.

[0035] In one embodiment, based on the computing resource scheduling policy chain that has been generated, the specific circumstances of the resource scheduling decisions of each task to be executed are analyzed, and by evaluating these decisions, it is checked whether the resource allocation is balanced to ensure that resources are reasonably allocated between local and cloud computing nodes, and the balancing coefficient corresponding to each resource scheduling decision is calculated; when this balancing coefficient is greater than or equal to the balancing threshold, the computing resource scheduling policy chain is directly used as a computing resource scheduling optimization chain, otherwise, the computing resource scheduling policy chain is optimized and scheduled to generate a computing resource scheduling optimization chain to avoid resource overload or idleness, improve task execution efficiency, and ensure that all computing nodes in the network receive reasonable load distribution.

[0036] Furthermore, the present application provides a method for optimizing the computing power balance of the local cloud collaborative network according to the computing resource scheduling strategy chain to obtain a computing resource scheduling optimization chain, including: According to the computing resource scheduling strategy chain and the local cloud computing resource space, the computing power allocation balance evaluation is performed on the local cloud collaborative network to obtain a computing power allocation balance coefficient; it is determined whether the computing power allocation balance coefficient is less than the computing power allocation balance threshold; if the computing power allocation balance coefficient is less than the computing power allocation balance threshold, the computing resource scheduling strategy chain is optimized and scheduled according to the computing power allocation balance threshold to generate the computing resource scheduling optimization chain.

[0037] Preferably, the resource allocation of the local cloud collaborative network is evaluated for computing power distribution balance according to the existing computing resource scheduling strategy chain and the local cloud computing resource space. The main task of the computing power distribution balance evaluation is to check whether the current resource scheduling scheme achieves the load balancing goal and ensure that the resource allocation is evenly distributed between the local computing resources and the cloud computing resources. When performing the computing power distribution balance evaluation, for each resource scheduling strategy in the computing resource scheduling strategy chain, the corresponding ratio calculation is performed between the resource demand involved and the available resources in the local computing resources and the cloud computing resources to obtain the local resource utilization rate and the cloud resource utilization rate involved. For example, If the resource scheduling strategy of a task to be executed is to allocate 3 CPU cores, 6GB memory, and 30GB storage from the local machine, and allocate 1 CPU core, 2GB memory, and 30GB storage from the cloud machine, and the local available CPU cores are 10, the memory is 64GB, and the available storage is 200GB, and the cloud available CPU cores are 20, the memory is 128GB, and the available storage is 500GB, then the local CPU core utilization rate of this resource scheduling strategy is 3 / 10, the local memory utilization rate is 6 / 64, the local storage utilization rate is 30 / 200, the cloud CPU core utilization rate is 1 / 20, and the cloud memory utilization rate is 2 / 128, the cloud storage usage rate is 30 / 500; then, the local resource usage rate is averaged and the cloud resource usage rate is averaged to obtain the average usage rate of local computing resources and the average usage rate of cloud computing resources, and then the absolute difference between the average usage rate of local computing resources and the average usage rate of cloud computing resources is subtracted from 1 to obtain the computing power allocation balance coefficient; after the computing power allocation balance coefficient is calculated, the computing power allocation balance coefficient is compared with the preset computing power allocation balance threshold to determine whether the current resource allocation is reasonable; if the computing power allocation balance coefficient is less than the threshold, it means that the resources are unevenly allocated, which may cause some resources to be unbalanced. If one resource pool is overloaded and other resource pools are idle, the resource allocation of the tasks will be readjusted to balance the load between the local and cloud resource pools. For example, if the cloud computing resources are used less, some tasks will be allocated to the cloud, and vice versa. This can make the computing resources in the local cloud collaborative network more evenly used, avoiding the situation where the resources of a certain node are over-occupied while other nodes are idle. After the optimization scheduling is completed, a computing resource scheduling optimization chain is generated, which includes the optimized resource scheduling strategy to ensure that resources are evenly distributed between the local and cloud, avoiding resource overload and waste.

[0038] Based on the local cloud collaborative network, the to-be-executed task chain is executed according to the computing resource scheduling optimization chain.

[0039] In one embodiment, after obtaining the computing resource scheduling optimization chain, it is synchronized to the local cloud collaborative network to ensure that the task chain to be executed is smoothly executed according to the optimal resource configuration; then, based on each optimized resource scheduling strategy in the computing resource scheduling optimization chain, combined with the actual resource situation in the local cloud collaborative network, the tasks to be executed are allocated and scheduled, and each task to be executed will be allocated resources between local computing resources and cloud computing resources according to the decision in the optimization chain, so as to ensure that the task can be executed efficiently; in this way, balanced scheduling can be achieved in the local and cloud resource pools to avoid overload or waste of resources, while maximizing resource utilization and task execution efficiency; finally, all tasks to be executed will be executed in sequence according to the optimal strategy defined in the computing resource scheduling optimization chain, ensuring that the task can be completed within the specified time and achieving the expected execution quality and efficiency.

[0040] In summary, the embodiments of the present application have at least the following technical effects: The embodiment of the present application receives tasks to be executed, and uses multiple priority evaluation factors to constrain the priorities of the tasks, thereby generating a chain of tasks to be executed; by collecting local cloud computing resource information, a computing resource space is established, and resources are allocated to the task chain according to a computing resource scheduling optimization algorithm, thereby obtaining a computing resource scheduling strategy chain; then, based on the needs of computing power balance optimization, the strategy chain is optimized and scheduled to generate a computing resource scheduling optimization chain; the task priority is evaluated through multiple priority evaluation factors, and the task execution order is optimized in combination with the dependencies of task execution, ensuring that the tasks can be efficiently executed according to the computing resource optimization strategy. These technical effects jointly solve the technical problems of unbalanced scheduling and low resource utilization between cloud and local computing resources, and achieve the technical effect of improving the collaborative work efficiency and resource utilization of cloud and local computing resources through priority constraints and computing resource scheduling optimization.

[0041] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0042] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0043] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A computer and cloud computing power optimization method based on multi-task collaboration, characterized in that: The method is applied to a local cloud collaborative network, the local cloud collaborative network includes a plurality of local computers and a cloud computing platform, and the method includes: Receiving Q tasks to be executed from the local cloud collaborative network, where Q is a positive integer greater than 1; Priority constraints are imposed on the Q tasks to be executed according to the multi-priority evaluation factors, and a task chain to be executed is established; Collecting the available computing resource information of the local cloud collaborative network and establishing a local cloud computing resource space; Based on the local cloud computing resource space, computing resources are allocated to the task chain to be executed according to a computing resource scheduling optimization algorithm to obtain a computing resource scheduling strategy chain; Performing a computing power balance optimization of the local cloud collaborative network according to the computing resource scheduling strategy chain to obtain a computing resource scheduling optimization chain; Based on the local cloud collaborative network, the to-be-executed task chain is executed according to the computing resource scheduling optimization chain.

2. The computer and cloud computing power optimization method based on multi-task collaboration as claimed in claim 1, characterized in that: The computing resource scheduling optimization algorithm includes: Inputting the first task to be executed in the task chain to be executed into the computing resource requirement parsing model to obtain the computing resources required by the first task; Taking the computing resources required by the first task as the resource scheduling target, making a resource scheduling decision on the local cloud computing resource space to obtain a first task resource scheduling decision domain; Based on the first task to be executed, optimizing the first task resource scheduling decision domain according to the computing resource scheduling optimization module to obtain a first computing resource scheduling strategy; The local cloud computing resource space is updated according to the first computing resource scheduling strategy, and the computing resource demand analysis model and the computing resource scheduling optimization module are combined to continue to perform computing resource scheduling optimization on the second to-be-executed task ... the Qth to-be-executed task in the to-be-executed task chain to obtain the second computing resource scheduling strategy ... the Qth computing resource scheduling strategy; The computing resource scheduling policy chain is constructed according to the first computing resource scheduling policy, the second computing resource scheduling policy, ... the Qth computing resource scheduling policy.

3. The computer and cloud computing power optimization method based on multi-task collaboration as claimed in claim 2, characterized in that: Based on the first task to be executed, optimizing the first task resource scheduling decision domain according to the computing resource scheduling optimization module to obtain a first computing resource scheduling strategy includes: The computing resource scheduling optimization module includes a resource scheduling prediction unit and a resource scheduling compatibility calculation unit; Based on the first task to be executed, setting a first task resource scheduling expectation, wherein the first task resource scheduling expectation includes an expected resource utilization rate, an expected task execution efficiency, and an expected task execution quality coefficient corresponding to the first task to be executed; Based on the resource scheduling prediction unit, performing expected optimization on the first task resource scheduling decision domain according to the first task resource scheduling expectation to obtain a first task resource scheduling optimization domain; According to the resource scheduling compatibility calculation unit, the first task resource scheduling optimization domain is optimized iteratively to generate the first computing resource scheduling strategy.

4. The computer and cloud computing power optimization method based on multi-task collaboration as claimed in claim 3, characterized in that: Based on the resource scheduling prediction unit, performing expected optimization on the first task resource scheduling decision domain according to the first task resource scheduling expectation to obtain the first task resource scheduling optimization domain, including: The resource scheduling prediction unit includes a resource utilization prediction model, a task execution efficiency prediction model and a task execution quality prediction model; Extracting a first resource scheduling decision according to the first task resource scheduling decision domain; Inputting the first resource scheduling decision and the first to-be-executed task into the resource scheduling prediction unit to obtain a first resource scheduling prediction result; Determining whether the first resource scheduling prediction result meets the first task resource scheduling expectation; If the first resource scheduling prediction result meets the first task resource scheduling expectation, adding the first resource scheduling decision to the first task resource scheduling optimization domain; According to the resource scheduling prediction unit and the first task resource scheduling expectation, the first task resource scheduling decision domain is continued to be optimized to construct the first task resource scheduling optimization domain.

5. The computer and cloud computing power optimization method based on multi-task collaboration as claimed in claim 3, characterized in that: According to the resource scheduling compatibility calculation unit, performing optimization iteration on the first task resource scheduling optimization domain to generate the first computing resource scheduling strategy includes: Extracting the nth resource scheduling decision according to the first task resource scheduling optimization domain, where n is a positive integer; Inputting the nth resource scheduling prediction result corresponding to the nth resource scheduling decision into the resource scheduling compatibility calculation unit to obtain the nth resource scheduling compatibility coefficient, wherein the resource scheduling compatibility calculation unit includes a resource utilization weight, a task execution efficiency weight, and a task execution quality weight; Extract the n+1th resource scheduling decision according to the first task resource scheduling optimization domain, and calculate the n+1th resource scheduling compatibility coefficient in combination with the resource scheduling compatibility calculation unit; According to the nth resource scheduling fit coefficient and the n+1th resource scheduling fit coefficient, performing current scheduling fit optimization on the nth resource scheduling decision and the n+1th resource scheduling decision to obtain a current optimal resource scheduling decision; The current optimal resource scheduling decision is iteratively optimized based on the resource scheduling compatibility calculation unit and the first task resource scheduling optimization domain until the number of iterative optimizations meets the iterative optimization number constraint, thereby obtaining the first computing resource scheduling strategy.

6. The computer and cloud computing power optimization method based on multi-task collaboration as claimed in claim 1, characterized in that: The computing power balancing optimization of the local cloud collaborative network is performed according to the computing resource scheduling strategy chain to obtain a computing resource scheduling optimization chain, including: Performing a computing power distribution balance evaluation on the local cloud collaborative network according to the computing resource scheduling strategy chain and the local cloud computing resource space to obtain a computing power distribution balance coefficient; Determine whether the computing power allocation balance coefficient is less than the computing power allocation balance threshold; If the computing power allocation balance coefficient is less than the computing power allocation balance threshold, the computing resource scheduling strategy chain is optimized and scheduled according to the computing power allocation balance threshold to generate the computing resource scheduling optimization chain.

7. The computer and cloud computing power optimization method based on multi-task collaboration as claimed in claim 1, characterized in that: Priority constraints are imposed on the Q tasks to be executed according to the multi-priority evaluation factors, and a task chain to be executed is established, including: The multi-priority evaluation factors include task urgency, task importance and task delay loss; Performing a priority evaluation on the Q tasks to be executed according to the multi-priority evaluation factor to obtain Q task execution priority coefficients; Prioritize the Q tasks to be executed according to the Q task execution priority coefficients to establish an initial task chain; Collect dependency information of the Q tasks to be executed and construct a task dependency graph; According to the task dependency graph, the initial task chain is optimized to generate the to-be-executed task chain.

8. The computer and cloud computing power optimization method based on multi-task collaboration as claimed in claim 7, characterized in that: The Q tasks to be executed are evaluated for priority according to the multi-priority evaluation factors to obtain Q task execution priority coefficients, including: Performing priority evaluation on the Q tasks to be executed according to the multi-priority evaluation factors to obtain Q task priority evaluation results, wherein each task priority evaluation result includes a task urgency coefficient, a task importance coefficient, and a task delay loss coefficient corresponding to each task to be executed; Performing weight allocation according to the multivariate priority evaluation factors, establishing a task priority evaluation weighted network, wherein the task priority evaluation weighted network includes a task urgency weight, a task importance weight, and a task delay loss weight; The priority evaluation results of the Q tasks are input into the task priority evaluation weighted network to obtain the execution priority coefficients of the Q tasks.

Images