Computing power resource matching method and device

Abstract

The invention relates to the technical field of big data, in particular to a computing power resource matching method and device. The method comprises the following steps: performing state identification on an obtained computing power resource use data set corresponding to at least one computing node by adopting a computing power resource competition model, obtaining a computing power resource competition state data set, and obtaining a scheduling suggestion data set; determining a scheduling optimization target according to the computing power resource use data set, the computing power resource competition state data set and the scheduling suggestion data set, and obtaining a first computing power resource allocation strategy set corresponding to the at least one computing node according to the scheduling optimization target and a genetic algorithm; and taking the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set as a target computing power resource allocation strategy, and matching the computing power resource of each computing node in the at least one computing node by adopting the target computing power resource allocation strategy. The method can improve the utilization rate of computing power resources, and improves the overall performance of the system.

Description

Technical Field

[0001] The present disclosure relates to the field of big data technology, and in particular to a computing resource matching method and device. Background Art

[0002] With the rapid development of information technology, cloud computing and big data technologies are increasingly widely used in all walks of life, and enterprises' demand for computing resources is gradually increasing. Therefore, how to efficiently allocate and utilize these computing resources has become a research goal to improve computing system performance and reduce operating costs. For example, a static allocation strategy can be used to allocate resources such as CPU, memory, and network bandwidth according to pre-set rules. Fixed thresholds and rules are used to identify and handle abnormal use of computing resources. In a multi-tenant environment, resource allocation and management mainly rely on manual adjustments and preset strategies. As a result, there will be low resource utilization and unintelligent allocation strategies, which will lead to a decrease in the overall performance of the system. Summary of the invention

[0003] The present disclosure provides a computing power resource matching method and device, which can improve the matching of computing power resource allocation strategy with the current allocation scenario, improve the intelligence of computing power resource allocation strategy determination, improve the utilization rate of computing power resources, and thus improve the overall performance of the system. The technical solution of the present disclosure is as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, a computing resource matching method is provided, including:

[0005] Using a computing power resource competition model to identify the state of the computing power resource usage data set corresponding to at least one computing node, and obtaining a computing power resource competition state data set;

[0006] Acquire a scheduling suggestion data set according to the computing power resource competition state data set;

[0007] Determine a scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set, and the scheduling suggestion data set, and obtain a first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and the genetic algorithm;

[0008] The computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set is used as the target computing power resource allocation strategy, and the computing power resources of each computing node in the at least one computing node are matched using the target computing power resource allocation strategy.

[0009] According to some embodiments, the method further comprises:

[0010] Obtain a collection of historical computing resource usage data through the database;

[0011] Preprocessing the historical computing resource usage data set to obtain a processed historical computing resource usage data set, wherein the preprocessing includes at least one of removing outliers, filling missing values, and data normalization;

[0012] Using an isolation forest algorithm to identify the processed historical computing power resource usage data set, and obtaining at least one abnormal data in the processed historical computing power resource usage data set;

[0013] Annotating the at least one abnormal data to obtain an annotated historical computing resource usage data set;

[0014] The annotated historical computing power resource usage data set and the Bayesian network algorithm are used to perform model training to obtain a computing power resource competition model.

[0015] According to some embodiments, determining a scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set, and the scheduling suggestion data set, and obtaining a first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and a genetic algorithm, includes:

[0016] Determining computing resource utilization and response time as scheduling optimization targets according to the computing resource usage data set, the computing resource competition status data set, and the scheduling suggestion data set;

[0017] According to the scheduling optimization objective and the genetic algorithm, a second computing power resource allocation strategy set corresponding to at least one computing node is obtained;

[0018] A multivariate linear regression model is used to analyze the system performance based on the historical resource utilization and the historical response time, and a first weight corresponding to the computing resource utilization and a second weight corresponding to the response time are obtained;

[0019] Obtaining, according to the computing power resource utilization rate, the response time, the first weight, the second weight, and the fitness function, a fitness value corresponding to each computing power resource allocation strategy in the second computing power resource allocation strategy set;

[0020] Obtain computing power resource allocation strategies whose fitness values ​​are greater than the fitness threshold in the second computing power resource allocation strategy set and add them to the third computing power resource allocation strategy set;

[0021] According to a preset mutation probability, cross-processing is performed on the third computing power resource allocation strategy set to obtain a fourth computing power resource allocation strategy set;

[0022] A genetic algorithm is used to optimize the fourth computing power resource allocation strategy set to obtain the first computing power resource allocation strategy set.

[0023] According to some embodiments, after matching the computing resources of each computing node in the at least one computing node using the target computing resource allocation strategy, the method further includes:

[0024] Obtaining a performance indicator of each computing node in the at least one computing node;

[0025] Obtaining a computing power resource allocation strategy effect evaluation report corresponding to each computing node according to the performance indicator;

[0026] According to the computing power resource allocation strategy effect evaluation report, the target computing power resource allocation strategy is adjusted to obtain an adjusted target computing power resource allocation strategy;

[0027] The computing resources of each computing node in the at least one computing node are adjusted using the adjusted target computing resource allocation strategy.

[0028] According to some embodiments, the adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes:

[0029] Acquire monitoring data corresponding to each computing node through a sensor, wherein the monitoring data includes a computing resource usage data set and a service quality indicator data set;

[0030] Using a target sensitivity classification model to identify the task set corresponding to each computing node, and obtaining a sensitivity classification result corresponding to each task in the task set;

[0031] When the monitoring data meets the task migration triggering condition, obtaining the computing node for task migration;

[0032] Determine, according to the current network topology information, the computing resource usage data of each computing node, and the sensitivity classification result, a migration task having a sensitivity greater than a sensitivity threshold;

[0033] The Dijkstra algorithm is used to calculate the shortest path, migration time and used resources from the source computing node to the target computing node, and determine the task migration strategy, wherein the target computing node is a node among the at least one computing node that meets the task migration requirements;

[0034] The task migration strategy is adopted to migrate the migration task from the source computing node to the target computing node.

[0035] According to some embodiments, the method further comprises:

[0036] According to the computing power resource usage data set of each computing node, the computing power resource usage data set is preprocessed to obtain a processed computing power resource usage data set;

[0037] Using a sliding window technology, the computing resource usage data of each tenant in the tenant set within the sliding window is obtained, wherein the tenant set corresponds to each computing node;

[0038] Get the average computing resource usage data in each time window;

[0039] According to the preset usage rate threshold set and the average computing resource usage data, a first time period and a second time period are acquired, wherein the first time period is a time period in which the usage rate is greater than a first preset usage rate threshold in the usage rate threshold set, and the second time period is a time period in which the usage rate is less than a second preset usage rate threshold in the usage rate threshold set, and the first usage rate threshold is greater than the second usage rate threshold;

[0040] Obtaining the computing power resource contention degree of each tenant in the first time period according to the computing power resource contention degree algorithm corresponding to the first time period;

[0041] The task migration strategy is adjusted according to the computing power resource competition degree to obtain an adjusted task migration strategy.

[0042] According to some embodiments, the adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes:

[0043] Call data based on the task computing resources corresponding to each computing node;

[0044] Using the Apriori algorithm and the K-means clustering algorithm, the dependency relationship between at least one task corresponding to each computing node is obtained;

[0045] According to the dependency relationship between the tasks and the processing time log of each task, obtaining the processing time of each task;

[0046] According to the processing time of each task, a task dependency data set is obtained, wherein the task dependency data set includes the execution time of the task, the execution time of the dependent tasks and the computing resource consumption index;

[0047] Acquire, according to the task dependency data set, a time sensitivity of a task dependency between the first task and a second task, wherein the first task is any task of at least one task, and the second task is any task of the at least one task except the first task;

[0048] Obtaining a task migration order according to the time sensitivity of the task dependencies;

[0049] The computing resources of each computing node in the at least one computing node are matched according to the target computing resource allocation strategy in the task migration order.

[0050] According to some embodiments, the adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes:

[0051] Using a computing power resource demand prediction model and a computing power resource usage data set to obtain computing power resource demand information corresponding to each computing task;

[0052] The computing power resources of each computing node in the at least one computing node are matched according to the computing power resource demand information and the target computing power resource allocation strategy.

[0053] According to a second aspect of an embodiment of the present disclosure, a computing power resource matching device is provided, including:

[0054] A set acquisition unit, used to identify the state of the acquired computing resource usage data set corresponding to at least one computing node using a computing resource competition model, and acquire a computing resource competition state data set;

[0055] The set acquisition unit is used to acquire a scheduling suggestion data set according to the computing power resource competition state data set;

[0056] a target determination unit, configured to determine a scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set, and the scheduling suggestion data set, and to obtain a first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and a genetic algorithm;

[0057] A resource matching unit is used to use the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set as the target computing power resource allocation strategy, and use the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node.

[0058] According to a third aspect of an embodiment of the present disclosure, a network device is provided, including:

[0059] processor;

[0060] a memory for storing instructions executable by the processor;

[0061] Among them, the processor is configured to execute the instructions to implement the computing power resource matching method described in any one of the aforementioned aspects.

[0062] According to a fourth aspect of an embodiment of the present disclosure, a storage medium is provided. When instructions in the storage medium are executed by a processor of a network device, the network device is enabled to execute the computing power resource matching method described in any one of the preceding aspects.

[0063] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program implements any one of the methods described in the preceding aspects when executed by a processor.

[0064] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0065] In some or related embodiments, by adopting a computing power resource competition model to identify the state of the computing power resource usage data set corresponding to at least one computing node, a computing power resource competition state data set is obtained; according to the computing power resource competition state data set, a scheduling recommendation data set is obtained; according to the computing power resource usage data set, the computing power resource competition state data set and the scheduling recommendation data set, a scheduling optimization target is determined, and according to the scheduling optimization target and the genetic algorithm, a first computing power resource allocation strategy set corresponding to at least one computing node is obtained; the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set is used as the target computing power resource allocation strategy, and the target computing power resource allocation strategy is used to match the computing power resources of each computing node in the at least one computing node. Therefore, the computing power resource allocation strategy can be dynamically adjusted through the computing power resource usage data, the matching of the computing power resource allocation strategy with the current allocation scenario can be improved, the computing power resource allocation strategy can be optimized, the situation where the competition for computing power resources makes the computing power resource utilization rate low can be reduced, the intelligence of the determination of the computing power resource allocation strategy can be improved, the utilization rate of computing power resources can be improved, and the overall performance of the system can be improved.

[0066] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0068] Figure 1 is a flow chart of a computing resource matching method according to an exemplary embodiment;

[0069] Figure 2 is a flow chart of a computing resource matching method according to an exemplary embodiment;

[0070] Figure 3 is a block diagram of a computing resource matching device according to an exemplary embodiment;

[0071] Figure 4 The invention is a block diagram of a network device according to an exemplary embodiment. DETAILED DESCRIPTION

[0072] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0073] The embodiments of the present disclosure propose a computing resource matching method and device. In some embodiments, the computing resource matching method and information processing method, communication method and other terms can be interchangeable, the computing resource matching device and information processing device, communication device and other terms can be interchangeable, and the information processing system and communication system and other terms can be interchangeable.

[0074] The embodiments of the present disclosure are not exhaustive, but are only illustrative of some embodiments, and are not intended to be a specific limitation on the scope of protection of the present disclosure. In the absence of contradiction, each step in a certain embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a certain embodiment can also be implemented as an independent embodiment, and the order of the steps in a certain embodiment can be arbitrarily exchanged. In addition, the optional implementation methods in a certain embodiment can be arbitrarily combined; in addition, the embodiments can be arbitrarily combined, for example, some or all of the steps of different embodiments can be arbitrarily combined, and a certain embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0075] In each embodiment of the present disclosure, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between the embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form a new embodiment based on their internal logical relationships.

[0076] The terms used in the embodiments of the present disclosure are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure.

[0077] In the embodiments of the present disclosure, unless otherwise specified, elements expressed in the singular form, such as "a", "an", "the", "above", "said", "aforementioned", "this", etc., may mean "one and only one", or "one or more", "at least one", etc. For example, when using articles such as "a", "an", "the" in English in translation, the noun after the article may be understood as a singular expression or a plural expression.

[0078] In the embodiments of the present disclosure, “plurality” refers to two or more.

[0079] In some embodiments, the terms “at least one,” “one or more,” “a plurality of,” “multiple,” etc. may be used interchangeably.

[0080] The prefixes such as "first" and "second" in the embodiments of the present disclosure are only used to distinguish different description objects, and do not constitute restrictions on the position, order, priority, quantity or content of the description objects. The statement of the description object refers to the description in the context of the claims or embodiments, and should not constitute unnecessary restrictions due to the use of prefixes. For example, if the description object is a "field", the ordinal number before the "field" in the "first field" and the "second field" does not limit the position or order between the "fields", and the "first" and "second" do not limit whether the "fields" they modify are in the same message, nor do they limit the order of the "first field" and the "second field". For another example, if the description object is a "level", the ordinal number before the "level" in the "first level" and the "second level" does not limit the priority between the "levels". For another example, the number of description objects is not limited by the ordinal number, and can be one or more. Taking the "first device" as an example, the number of "devices" can be one or more. In addition, the objects modified by different prefixes may be the same or different. For example, if the description object is "device", then the "first device" and the "second device" may be the same device or different devices, and their types may be the same or different. For another example, if the description object is "information", then the "first information" and the "second information" may be the same information or different information, and their contents may be the same or different.

[0081] In some embodiments, "terminal" or "terminal device" can be referred to as "user equipment (UE)", "user terminal" "mobile station (MS)", "mobile terminal (MT)", subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, etc.

[0082] In some embodiments, data, information, etc. may be obtained with the user's consent.

[0083] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0084] Figure 1 is a flowchart of a computing resource matching method according to an exemplary embodiment. Figure 1 As shown, the computing power resource matching method can be used in a computing power matching scenario based on big data, including the following steps:

[0085] In step S11, a computing resource competition model is used to identify the state of the computing resource usage data set corresponding to at least one computing node, and a computing resource competition state data set is obtained;

[0086] According to some embodiments, the execution subject of the embodiments of the present disclosure is a network device. The name of the network device is not limited. For example, the network device can also be called a server, or a server cluster, etc.

[0087] According to some embodiments, the computing power resource competition model may be, for example, a model that has been trained and can be used to determine the computing power resource competition state. The computing power resource competition model does not specifically refer to a fixed model. For example, when the model parameters corresponding to the computing power resource competition model change, the computing power resource competition model may also change accordingly. For example, when the model type corresponding to the computing power resource competition model changes, the computing power resource competition model may also change accordingly.

[0088] In some embodiments, the computing node may be, for example, a node for calculating computing power. The name of the settlement node is not limited. For example, the computing node may also be called a computing server, a computing device, etc. The number of computing nodes may be, for example, at least one, that is, the number of computing nodes is not limited.

[0089] In some embodiments, the computing resource usage data set may include, for example, a collection of at least one computing resource usage data. The computing resource usage data set does not specifically refer to a fixed set. For example, when the amount of data corresponding to the computing resource usage data set changes, the computing resource usage data may also change accordingly. For example, when any computing resource usage data in the computing resource usage data set changes, the computing resource usage data may also change accordingly.

[0090] According to some embodiments, the state identification may be used, for example, to identify the current computing power resource competition state. The computing power resource competition state may be used, for example, to indicate whether there is a competition relationship between computing power resources.

[0091] In some embodiments, the computing power resource competition state data set may include, for example, a collection of at least one computing power resource competition state data. The computing power resource competition state data does not specifically refer to a fixed set. For example, when the number of state data included in the computing power resource competition state data set changes, the computing power resource competition state data set may also change accordingly. For example, when a certain state data in the computing power resource competition state data set changes, the computing power resource competition state data set may also change accordingly.

[0092] In some embodiments, a computing power resource competition model is used to identify the state of a computing power resource usage data set corresponding to at least one computing node, and obtain a computing power resource competition state data set.

[0093] In step S12, a scheduling suggestion data set is obtained according to the computing power resource competition state data set;

[0094] According to some embodiments, the scheduling suggestion data set may be, for example, a set formed by gathering at least one scheduling suggestion data. The name of the scheduling suggestion data is not limited. For example, the scheduling suggestion data may also be called a scheduling strategy data set, or a scheduling mode set, etc.

[0095] In some embodiments, when the computing power resource competition state data set is obtained, a scheduling suggestion data set may be obtained based on the computing power resource competition state data set. The method for obtaining the scheduling suggestion data set is not limited.

[0096] In step S13, a scheduling optimization target is determined according to the computing power resource usage data set, the computing power resource competition state data set and the scheduling suggestion data set, and a first computing power resource allocation strategy set corresponding to at least one computing node is obtained according to the scheduling optimization target and the genetic algorithm;

[0097] According to some embodiments, the scheduling optimization target can be used to indicate a target that needs to be optimized. The scheduling optimization target does not specifically refer to a fixed target. For example, when the number of targets corresponding to the scheduling optimization target changes, the scheduling optimization target can also change accordingly. For example, when a specific target corresponding to the scheduling optimization target occurs, the scheduling optimization target can also change accordingly. Among them, the scheduling optimization target can, for example, include at least one of computing power resource utilization and response time. For example, the scheduling optimization target can, for example, include computing power resource utilization and response time.

[0098] In some embodiments, the genetic algorithm may be, for example, an algorithm that has been determined to generate a computing resource allocation strategy. The genetic algorithm does not specifically refer to a fixed algorithm, for example, when a parameter in the genetic algorithm changes, the genetic algorithm may also change accordingly.

[0099] According to some embodiments, the computing power resource allocation strategy set may be, for example, a group formed by at least one computing power resource allocation strategy. The computing power resource allocation strategy set does not specifically refer to a fixed set. For example, when the number of strategies corresponding to the computing power resource allocation strategy set changes, the computing power resource allocation strategy set may also change accordingly. For example, when any computing power resource allocation strategy in the computing power resource allocation strategy set changes, the computing power resource allocation strategy set may also change accordingly.

[0100] According to some embodiments, the first computing resource allocation strategy set is used, for example, to indicate that a computing resource allocation strategy set corresponding to at least one computing node is obtained according to the scheduling optimization objective and the genetic algorithm. The first in the first computing resource allocation strategy set is only used to distinguish from the remaining computing resource allocation strategy sets, and does not specifically refer to a fixed computing resource allocation strategy set.

[0101] In some embodiments, for example, a scheduling optimization target can be determined based on the computing power resource usage data set, the computing power resource competition status data set, and the scheduling recommendation data set, and a first computing power resource allocation strategy set corresponding to at least one computing node can be obtained based on the scheduling optimization target and a genetic algorithm.

[0102] In step S14, the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set is used as the target computing power resource allocation strategy, and the computing power resources of each computing node in the at least one computing node are matched using the target computing power resource allocation strategy.

[0103] According to some embodiments, the fitness value can be used to indicate the ability of the computing power resource allocation strategy to affect the system performance, that is, it can be used to indicate whether the computing power resource allocation strategy is suitable for the current system. The fitness value does not specifically refer to a fixed value. For example, when the method for determining the fitness value changes, the fitness value may also change accordingly. For example, when the computing power resource allocation strategy changes, the fitness value corresponding to the computing power resource allocation strategy may also change accordingly.

[0104] In some embodiments, the target computing power resource allocation strategy may be, for example, a computing power resource allocation strategy determined from a plurality of computing power resource allocation strategies, and the target computing power resource allocation strategy does not specifically refer to a fixed computing power resource allocation strategy, for example, when the first computing power resource allocation strategy set changes, the target computing power resource allocation strategy may also change accordingly. For example, when the specific resource allocation information corresponding to the target computing power resource allocation strategy changes, the target computing power resource allocation strategy may also change accordingly.

[0105] According to some embodiments, the computing power resources of each computing node can be used to indicate the computing power resource information corresponding to each computing node. The computing power resources of each node do not specifically refer to a fixed resource. For example, when a computing power resource allocation instruction for each computing node is received, the computing power resources of each computing node may also change accordingly. For example, when a computing node changes, the computing power resources of each computing node may also change accordingly. The computing power resources may, for example, refer to various computing capabilities required in the computing process, including CPU, memory, network bandwidth, etc.

[0106] In some embodiments, the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set is used as the target computing power resource allocation strategy, and the target computing power resource allocation strategy is used to match the computing power resources of each computing node in the at least one computing node.

[0107] In some or related embodiments, a computing power resource competition model is used to perform state identification on a computing power resource usage data set corresponding to at least one computing node, thereby obtaining a computing power resource competition state data set; based on the computing power resource competition state data set, a scheduling recommendation data set is obtained; based on the computing power resource usage data set, the computing power resource competition state data set and the scheduling recommendation data set, a scheduling optimization target is determined, and based on the scheduling optimization target and a genetic algorithm, a first computing power resource allocation strategy set corresponding to at least one computing node is obtained; the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set is used as the target computing power resource allocation strategy, and the target computing power resource allocation strategy is used to match the computing power resources of each computing node in the at least one computing node. Therefore, the computing resource allocation strategy can be dynamically adjusted through computing resource usage data, which can improve the matching of computing resource allocation strategy with the current allocation scenario or resource demand, and can optimize the computing resource allocation strategy to reduce the competition for computing resources, which leads to low computing resource utilization. The intelligence of computing resource allocation strategy determination can be improved, and the utilization rate of computing resources can be improved, thereby improving the overall performance of the system. In addition, the competition status of computing resources can be judged and processed, the optimization time of optimizing computing resource allocation strategy can be reduced, the service quality and user experience of tenants can be improved, and the operating costs can be reduced.

[0108] Figure 2 is a flowchart of a computing resource matching method according to an exemplary embodiment. Figure 2 As shown, the computing power resource matching method can be used in a scenario where tasks are migrated between computing nodes based on computing power resources, and includes the following steps:

[0109] In step S21, a computing resource competition model is used to identify the state of the computing resource usage data set corresponding to at least one computing node, and a computing resource competition state data set is obtained;

[0110] According to some embodiments, the execution subject of the embodiments of the present disclosure is a network device. The name of the network device is not limited. For example, the network device can also be called a server, or a server cluster, etc.

[0111] In some embodiments, for example, a computing resource usage data set corresponding to at least one computing node may be obtained by a sensor at a preset time interval. The preset time interval does not specifically refer to a fixed time interval. For example, when a modification instruction for the preset time interval is received, the preset time interval may also change accordingly. For example, when the duration corresponding to the preset time interval changes, the preset time interval may also change accordingly. For example, when the current usage scenario changes, the preset time interval may also change accordingly.

[0112] In some embodiments, the sensor does not specifically refer to a fixed sensor. For example, when the computing node changes, the sensor may also change accordingly. For example, when the acquisition accuracy corresponding to the computing resource usage data changes, the sensor may also change accordingly.

[0113] In some embodiments, the computing power resource usage data set may also be stored in a database, where the computing power resource usage data set may include, for example, CPU, memory, network bandwidth, etc. The computing power resource usage data set does not specifically refer to a fixed set. For example, when the amount of computing power resource usage data corresponding to the computing power resource usage data set changes, the computing power resource usage data set may also change accordingly.

[0114] According to some embodiments, the method further comprises:

[0115] Obtain a collection of historical computing resource usage data through the database;

[0116] Preprocessing the historical computing resource usage data set to obtain a processed historical computing resource usage data set, wherein the preprocessing includes at least one of removing outliers, filling missing values, and data normalization;

[0117] Using an isolation forest algorithm to identify the processed historical computing power resource usage data set, and obtaining at least one abnormal data in the processed historical computing power resource usage data set;

[0118] Annotating the at least one abnormal data to obtain an annotated historical computing resource usage data set;

[0119] The annotated historical computing resource usage data set and the Bayesian network algorithm are used for model training to obtain a computing resource competition model. Therefore, the data can be preprocessed to improve the accuracy of computing resource usage data acquisition, reduce the impact of redundant data on computing resource allocation, and improve the accuracy of computing resource allocation.

[0120] In some embodiments, the preprocessing of the historical computing resource usage data set may also include other processing methods, and may not be limited to the processing method shown in the embodiment of the present disclosure.

[0121] According to some embodiments, the isolation forest algorithm does not specifically refer to a fixed algorithm. For example, when the parameters corresponding to the isolation forest algorithm change, the isolation forest algorithm may also change accordingly. Among them, abnormal data may be, for example, data that does not meet the requirements. The abnormal data does not specifically refer to a fixed data. For example, when the historical computing power resource usage data set changes, the abnormal data may also change accordingly. For example, when the preprocessing of the historical computing power resource usage data set changes, the abnormal data may also change accordingly.

[0122] In some embodiments, the marking of at least one abnormal data does not specifically refer to a fixed marking method. For example, the marking method may be a binary marking method. The historical computing power resource usage data may be obtained through a database, and when the abnormal data is obtained, the historical computing power usage data set may be marked for computing power resource competition using a binary marking method. For example, it may be marked as 1 when there is resource competition, and marked as 0 when there is no resource competition. Or other marking numbers, that is, 0 and 1 are only for illustration. Among them, the marking method may also be composed of different information, for example.

[0123] In some embodiments, the computing power resource competition model may be, for example, a model that has been trained to identify the computing power resource competition state. The computing power resource competition model does not specifically refer to a fixed model. For example, when the historical computing power resource usage data set changes, the computing power resource competition model may also change accordingly. For example, when the model conditions corresponding to the computing power resource competition model change, the computing power resource competition model may also change accordingly.

[0124] In some embodiments, when the computing power resource competition model is obtained, the computing power resource competition model can be used to identify the state of the computing power resource usage data set corresponding to the obtained at least one computing node, and obtain the computing power resource competition state data set. Among them, the computing power resource competition state data can include existence or non-existence. Among them, the computing power resource competition state can be applied to the computing power resource competition state of the current computing node, for example.

[0125] In step S22, a scheduling suggestion data set is obtained according to the computing power resource competition state data set;

[0126] The specific process is as described above and will not be repeated here.

[0127] In step S23, a scheduling optimization target is determined according to the computing power resource usage data set, the computing power resource competition state data set and the scheduling suggestion data set, and a first computing power resource allocation strategy set corresponding to at least one computing node is obtained according to the scheduling optimization target and the genetic algorithm;

[0128] The specific process is as described above and will not be repeated here.

[0129] In some embodiments, for example, based on the scheduling suggestion data set, the real-time computing resource usage data set, and the real-time computing resource competition status data set, computing resource utilization and response time can be determined as scheduling optimization targets, and a genetic algorithm is used to generate N random computing resource allocation strategies to form an initial population. Wherein, N is a positive integer, indicating the number of computing resource allocation strategies included in the first computing resource allocation strategy set.

[0130] According to some embodiments, determining a scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set, and the scheduling suggestion data set, and obtaining a first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and a genetic algorithm, includes:

[0131] Determining computing resource utilization and response time as scheduling optimization targets according to the computing resource usage data set, the computing resource competition status data set, and the scheduling suggestion data set;

[0132] According to the scheduling optimization objective and the genetic algorithm, a second computing power resource allocation strategy set corresponding to at least one computing node is obtained;

[0133] A multivariate linear regression model is used to analyze the system performance based on the historical resource utilization and the historical response time, and a first weight corresponding to the computing resource utilization and a second weight corresponding to the response time are obtained;

[0134] Obtaining, according to the computing power resource utilization rate, the response time, the first weight, the second weight, and the fitness function, a fitness value corresponding to each computing power resource allocation strategy in the second computing power resource allocation strategy set;

[0135] Obtain computing power resource allocation strategies whose fitness values ​​are greater than the fitness threshold in the second computing power resource allocation strategy set and add them to the third computing power resource allocation strategy set;

[0136] According to a preset mutation probability, cross-processing is performed on the third computing power resource allocation strategy set to obtain a fourth computing power resource allocation strategy set;

[0137] A genetic algorithm is used to optimize the fourth computing power resource allocation strategy set to obtain the first computing power resource allocation strategy set.

[0138] According to some embodiments, the fitness function may be, for example, Fitness = α1 / computing resource utilization + β response time. When resource utilization and response time are used as evaluation indicators, the fitness value of each computing resource allocation strategy may be calculated, wherein α and β are weight parameters, α may be, for example, the first weight, and β may be, for example, the second weight. In other words, the weight parameter does not specifically refer to a fixed parameter. For example, when a modification instruction for the weight parameter is received, the weight parameter may also change accordingly. For example, when one of the weight parameters changes, the other weight parameter may also change accordingly.

[0139] In some embodiments, a multivariate linear regression model is used to analyze the impact of resource utilization and response time on system performance based on historical data of resource utilization and response time, wherein α and β can be determined by minimizing the regression error. The computing power resource allocation strategies with fitness values ​​greater than the preset fitness index are selected to form a breeding pool. Among them, the crossover processing of the embodiment of the present disclosure does not specifically refer to a fixed processing. For example, the computing power resource allocation strategies in the breeding pool can be used to generate new computing power resource allocation strategies using crossover processing methods such as single-point crossover, two-point crossover or uniform crossover, and the new computing power resource allocation strategies are mutated with a preset mutation probability to increase the diversity of the population. Repeating the selection, crossover and mutation process, the genetic algorithm can be run according to the set number of iterations or fitness threshold, and the computing power resource allocation strategy can be gradually optimized, and the computing power resource allocation strategy with the highest fitness value can be selected as the optimal computing power resource allocation strategy. For example, the acquisition process for the first computing power resource allocation strategy set can be re-executed multiple times until the first computing power resource allocation strategy set that meets the requirements is obtained.

[0140] In step S24, the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set is used as the target computing power resource allocation strategy, and the computing power resources of each computing node in the at least one computing node are matched by using the target computing power resource allocation strategy;

[0141] The specific process is as described above and will not be repeated here.

[0142] According to some embodiments, the adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes:

[0143] Acquire monitoring data corresponding to each computing node through a sensor, wherein the monitoring data includes a computing resource usage data set and a service quality indicator data set;

[0144] Using a target sensitivity classification model to identify the task set corresponding to each computing node, and obtaining a sensitivity classification result corresponding to each task in the task set;

[0145] When the monitoring data meets the task migration triggering condition, obtaining the computing node for task migration;

[0146] Determine, according to the current network topology information, the computing resource usage data of each computing node, and the sensitivity classification result, a migration task having a sensitivity greater than a sensitivity threshold;

[0147] The Dijkstra algorithm is used to calculate the shortest path, migration time and used resources from the source computing node to the target computing node, and determine the task migration strategy, wherein the target computing node is a node among the at least one computing node that meets the task migration requirements;

[0148] The task migration strategy is adopted to migrate the migration task from the source computing node to the target computing node. Therefore, it is possible to determine whether to perform task migration through monitoring data, improve the execution success rate of the migrated task, and perform the migration timing of computing resources according to the migration strategy, improve the accuracy of computing resource allocation, reduce the waste of computing resources and service interruption, and improve the overall performance of the system.

[0149] According to some embodiments, the service quality indicator data set may be, for example, a collection of at least one service quality indicator data. The service quality indicator data set does not specifically refer to a fixed set. For example, when the amount of data corresponding to the service quality indicator data set changes, the service quality indicator data set may also change accordingly. For example, when a certain service quality indicator data in the service quality indicator data set changes, the service quality indicator data set may also change accordingly. Among them, the service quality indicator data set may include, for example, response time and success rate.

[0150] According to some embodiments, a hierarchical analysis method can be used based on a service quality indicator data set. The defined evaluation indicators may include response time, success rate, and computing resource utilization rate. A hierarchical model is constructed, including a target layer (service quality optimization), a criterion layer (various service quality indicators), and a solution layer (different resource allocation strategies). Based on expert experience or historical data, a relative importance matrix of each evaluation indicator is established.

[0151] The weight of each evaluation indicator is calculated through the eigenvalue method or the weighted average method. According to the weight of the evaluation indicator and the actual data, the comprehensive evaluation score is calculated and the computing power resource allocation adjustment strategy is formulated.

[0152] Through sensors, the computing resource usage data and computing resource competition status of each tenant are obtained. Based on the computing resource competition feedback mechanism, the computing resource allocation strategy between tenants is dynamically adjusted, including adjusting the resource quota of containers or virtual machines, such as CPU and memory limits, or adjusting resource scheduling strategies, such as the resource scheduling strategy in Kubernetes.

[0153] According to some embodiments, the computing resource usage data set and the service quality indicator data set may be stored in a database to form monitoring data. The monitoring data may also be referred to as a monitoring data set or a real-time monitoring data set.

[0154] In some embodiments, the real-time monitoring data set is compared with the preset task migration trigger condition parameters to determine whether the real-time monitoring data meets the task migration trigger condition, wherein the preset task migration trigger condition may include, for example, a preset computing power resource competition degree threshold and a preset response time threshold. If the computing power resource competition degree and the response time exceed the set thresholds, a migration trigger signal is output.

[0155] According to some embodiments, the real-time monitoring data set can be segmented according to preset time windows to form time series data, and basic statistical features can be extracted from the time series data to form a feature data set. The basic statistical features include mean value, standard deviation, maximum value, minimum value, and fluctuation range.

[0156] According to some embodiments, the feature data set can be divided into a training set and a test set in a ratio of 8:2, and hyperparameters such as learning rate, number of hidden layers, and batch size are adjusted. The RNN model is trained using the training set, and the number of training epochs is set to 100. The mean square error MSE is used as the loss function, and the Adam optimizer is used for optimization. A sensitivity classification model is constructed, and the model performance is evaluated using a test set. Indicators such as accuracy, recall rate, and F1 score are calculated. The generalization ability of the model is further verified by a cross-validation method to ensure its stability on different data sets. The latest data in the real-time monitoring data set is obtained, and each task is classified according to the trained sensitivity classification model, and the sensitivity of the task to different computing resources is marked.

[0157] According to some embodiments, for example, a computing node that needs to perform task migration may be identified based on a migration trigger signal.

[0158] According to some embodiments, the current network topology information and computing power resource usage data of the computing nodes can be obtained, and based on the task sensitivity classification results, computing nodes with high sensitivity, that is, sensitivity greater than the sensitivity threshold, are determined, and target nodes with sufficient computing resources and good network bandwidth are selected, and high-sensitivity tasks are preferentially migrated to nodes that meet the above standards, and the Dijkstra algorithm is used to calculate the shortest path from the source node to the target node, optimize the migration time and resource usage, and determine the task migration strategy. The task migration strategy includes determining the best migration path and target node, and the network topology information includes the connection relationship between nodes and the network bandwidth. Among them, the target node can be, for example, a node whose computing resources are greater than the resource threshold and whose network bandwidth meets the bandwidth requirements.

[0159] According to some embodiments, the current computing resource usage data and network topology information of the computing node can be obtained through sensors, and a migration plan can be formulated according to the task migration strategy, including the migrated tasks, migration paths, target nodes, and estimated migration time. According to the migration plan, the task migration is implemented, the migrated tasks are transferred from the source node to the target node, and the computing resource usage data and service quality indicators during the migration process are monitored in real time. By comparing the computing resource usage data and service quality indicator data before and after the task migration, the effectiveness of the task migration strategy is evaluated. If the improvement ratio of the computing resource usage data and the service quality indicator data is lower than the preset ratio threshold, it is judged that the effectiveness of the task migration strategy does not meet the standard, and the task migration strategy that does not meet the effectiveness is further optimized, and the migration trigger condition parameters such as the computing resource competition threshold and the response time threshold are adjusted, and the weight parameters of the task sensitivity classification model are adjusted to obtain the optimized task migration strategy.

[0160] According to some embodiments, the method further comprises:

[0161] According to the computing power resource usage data set of each computing node, the computing power resource usage data set is preprocessed to obtain a processed computing power resource usage data set;

[0162] Using a sliding window technology, the computing resource usage data of each tenant in the tenant set within the sliding window is obtained, wherein the tenant set corresponds to each computing node;

[0163] Get the average computing resource usage data in each time window;

[0164] According to the preset usage rate threshold set and the average computing resource usage data, a first time period and a second time period are acquired, wherein the first time period is a time period in which the usage rate is greater than a first preset usage rate threshold in the usage rate threshold set, and the second time period is a time period in which the usage rate is less than a second preset usage rate threshold in the usage rate threshold set, and the first usage rate threshold is greater than the second usage rate threshold;

[0165] Obtaining the computing power resource contention degree of each tenant in the first time period according to the computing power resource contention degree algorithm corresponding to the first time period;

[0166] The task migration strategy is adjusted according to the computing power resource competition degree to obtain an adjusted task migration strategy.

[0167] According to some embodiments, the acquisition of the computing power resource contention of each tenant in the first time period may be, for example, obtaining the real-time computing power resource usage data of the computing node through a sensor at a preset data acquisition frequency, and performing data preprocessing on the acquired real-time computing power resource usage data. The sliding window technology is used to count the computing power resource usage data of each tenant in the sliding window, calculate the average computing power resource usage rate in each time window, and identify the time periods with high usage and low usage based on the preset usage threshold. For the time window with high usage, the computing power resource contention formula (1) is used:

[0168]

[0169] The computing resource competition degree of all tenants in this time window can be calculated, where U ix represents the CPU usage of the ith tenant in the time window, U iy represents the memory usage of the ith tenant in the time window, U iz represents the network bandwidth usage of the ith tenant in the time window, N represents the total number of tenants, and C x Indicates the total CPU resource capacity, C y Indicates the total memory resource capacity, C z Indicates the total network bandwidth capacity.

[0170] Therefore, the computing power resources can be reallocated to each computing node through the scheduling system, and the computing power resource usage data and service quality indicators during the adjustment process can be recorded in real time. The performance indicators before and after the computing power resource allocation adjustment can be compared through the data analysis tool Pandas, and comparison charts and statistical data can be generated to evaluate the effect of the computing power resource allocation strategy, such as whether the response time is shortened, whether the success rate is improved, and whether the resource usage is more balanced; according to the effect of the computing power resource allocation strategy, the optimization direction of the computing power resource allocation strategy can be determined, and the computing power resource allocation strategy parameters can be adjusted.

[0171] According to some embodiments, the adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes:

[0172] Call data based on the task computing resources corresponding to each computing node;

[0173] Using the Apriori algorithm and the K-means clustering algorithm, the dependency relationship between at least one task corresponding to each computing node is obtained;

[0174] According to the dependency relationship between the tasks and the processing time log of each task, obtaining the processing time of each task;

[0175] According to the processing time of each task, a task dependency data set is obtained, wherein the task dependency data set includes the execution time of the task, the execution time of the dependent tasks and the computing resource consumption index;

[0176] Acquire, according to the task dependency data set, a time sensitivity of a task dependency between the first task and a second task, wherein the first task is any task of at least one task, and the second task is any task of the at least one task except the first task;

[0177] Obtaining a task migration order according to the time sensitivity of the task dependencies;

[0178] The computing resources of each computing node in the at least one computing node are matched according to the target computing resource allocation strategy in the task migration order. Therefore, the tenant tasks are migrated in sequence according to the task migration order to ensure the continuity and efficiency of task execution. The use of the time sensitivity formula can dynamically evaluate the time sensitivity of tasks, ensure the timely migration of high-priority tasks, improve the overall performance and stability of the system, and comprehensively consider the execution time and computing resource consumption of tasks when determining the task migration order, so that the migration decision is more scientific and reduces conflicts and waste in resource use. Through scientific determination of the migration order and maintenance of dependencies, the continuity and stability of the enterprise business are guaranteed, and business interruptions and performance degradation caused by task migration are reduced.

[0179] According to some embodiments, the adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes:

[0180] Using a computing power resource demand prediction model and a computing power resource usage data set to obtain computing power resource demand information corresponding to each computing task;

[0181] The computing power resources of each computing node in the at least one computing node are matched according to the computing power resource demand information and the target computing power resource allocation strategy.

[0182] According to some embodiments, for example, the real-time computing resource usage data and service quality indicator data of enterprise tenant A can be obtained through the hardware monitoring module Intel's RAPL and NetFlow sensors installed on the server and network equipment, and the computing resource usage data and service quality indicator data of enterprise tenant A can be obtained, for example, the CPU usage is 8500MHz, occupying 85% of the total CPU resources; the memory usage is 28GB, occupying 70% of the total memory resources; the network bandwidth usage is 600Mbps, occupying 60% of the total network bandwidth; the response time is 250 milliseconds, and the success rate is 95%. The preset task migration trigger condition parameters are the CPU usage threshold of 8000MHz, occupying 80% of the total CPU resources; the response time threshold is 200 milliseconds.

[0183] According to some embodiments, by comparing the real-time data of enterprise tenant A with the preset task migration trigger conditions, it is determined that its CPU usage and response time have exceeded the preset thresholds, indicating that enterprise tenant A meets the task migration trigger conditions and needs to perform task migration to optimize system performance. After confirming the need for task migration, the task computing resource call data of enterprise tenant A is further obtained through sensors, including an average task loading time of 10 seconds, a call frequency of 100 times per minute, a CPU usage of 1700MHz per call, a memory usage of 2GB per call, and a network bandwidth usage of 30Mbps per call. The acquired task computing resource call data is preprocessed, including data cleaning, format standardization and normalization, and the preprocessed task computing resource call data is obtained as follows: CPU usage is 0.20, memory usage is 0.10, and network bandwidth usage is 0.05.

[0184] According to some embodiments, the pre-processed task computing resource call data is converted into a transaction form, and each transaction represents a task call and its related resource consumption. For example, the task call frequency data can be converted into a transaction list, where each transaction contains the called task ID and the corresponding resource consumption (CPU, memory, network bandwidth). The Apriori algorithm is used for model training. For example, the support can be set to 0.1 and the confidence threshold can be set to 0.6. The transaction data set is scanned to find frequent item sets that meet the support threshold as candidate sets of task dependencies. Association rules are generated from the frequent item sets, the confidence of each rule is calculated, and the rules that meet the confidence threshold are retained. The generated association rules represent the call patterns and correlations between tasks. For example, the rule Task 1-Task 2 indicates that when Task 1 is called, Task 2 is also likely to be called, and the resource consumption patterns of the two are related. After the Apriori algorithm, the following candidate task dependency sets are obtained: the call frequency between task 1 and task 2 is 50 times per minute, and the resource sharing degree is high; the call frequency between task 1 and task 3 is 30 times per minute, and the resource sharing degree is medium; the call frequency between task 2 and task 4 is 40 times per minute, and the resource sharing degree is high; the call frequency between task 3 and task 4 is 20 times per minute, and the resource sharing degree is low. The candidate task dependency sets are converted into feature vectors, and each task pair is represented as a vector, which contains call frequency, time series, and resource sharing data. Task pair (task 1, task 2): [50, high]; task pair (task 1, task 3): [30, medium]; task pair (task 2, task 4): [40, high]; task pair (task 3, task 4): [20, low]. Set the number of clusters K = 3, initialize K cluster centers, randomly select K task pairs as initial centers, calculate the distance (Euclidean distance) between each task pair and the K cluster centers, assign the task pairs to the nearest cluster center, recalculate the center of each cluster, update the position of the cluster center, and repeat the above steps until the cluster center no longer changes or the maximum number of iterations is reached. The clustering results divide the task dependencies into 3 groups, Group 1: {Task 1, Task 2} and {Task 2, Task 4}, Group 2: {Task 1, Task 3}, Group 3: {Task 3, Task 4}. The task pairs in each group show high consistency in calling patterns and resource consumption. Based on the task dependency grouping results, obtain the task call frequency in each task dependency group, and according to the dependency strength coefficient formula:

[0185]

[0186] Calculate the dependency strength coefficient, where Aij is the number of times task i calls task j, Bi is the total number of times task i calls all tasks, task 1 calls task 2 30 times, and task 1 calls all tasks 50 times, then the dependency strength coefficient between task 1 and task 2 is 0.6. The above method takes into account the complex dependencies between tasks and real-time resource usage for the task migration problem in the enterprise environment. Unlike traditional task migration technology, this method obtains detailed task computing resource call data through sensors, and uses the Apriori algorithm and K-means clustering algorithm to accurately identify the dependencies and resource sharing patterns between tasks. This method ensures that the dependencies of tasks are fully considered during the task migration process, avoiding task execution failures caused by improper resource allocation or unsatisfied dependencies after migration. Through accurate task dependency grouping and strength coefficient calculation, the stability and business continuity of task migration are guaranteed, and the computing resource utilization and business processing efficiency of the enterprise are significantly improved.

[0187] According to some embodiments, based on the dependency relationship between tasks and the processing time log of each task, it can be obtained that the execution time of task A is 10 seconds, the execution time of task B is 12 seconds, and the execution time of task C is 8 seconds. A task dependency data set is created, and the task dependency data set includes the execution time of the tasks, the execution time of the dependent tasks, and the computing resource consumption index.

[0188] According to the execution time of the task, the execution time of the dependent tasks and the computing resource consumption index, the time sensitivity formula (2) is constructed:

[0189]

[0190] If the total execution time Ttotal is 30 seconds, the change ΔTB of the execution time of task B to task A is 2 seconds, and the change ΔTC of the execution time of task C to task B is 1.5 seconds.

[0191] According to some embodiments, the time sensitivity of a task dependency may be calculated using a time sensitivity formula, where the time sensitivity of a task dependency refers to the sensitivity of a task to changes in the execution time of its dependent tasks during execution.

[0192] Specifically, there may be a certain relationship between the execution time of a task and the execution time of other tasks it depends on. The execution time of a task with higher time sensitivity is very sensitive to the change of the execution time of the dependent task, while the execution time of a task with lower time sensitivity is less sensitive to the change of the execution time of the dependent task. For task A, the time sensitivity M(A) = 2 / 30 × 0.8 + 1.5 / 30 × 0.78 = 0.0923, M(B) = 0.06, M(C) = 0.0425. The migration order of tasks is determined according to the time sensitivity results of task dependencies. Since task A has the highest time sensitivity, task A is migrated first, followed by task B and task C.

[0193] According to some embodiments, the historical computing resource usage data of each tenant is obtained through sensors, and the historical computing resource usage data is preprocessed. Based on the preprocessed computing resource usage data, the long short-term memory network algorithm is used to perform model training and build a computing resource demand prediction model. Based on the real-time computing resource usage data, the computing resource demand prediction model is used to predict the computing resource demand of each tenant. Based on the computing resource demand prediction results, combined with the current computing resource data and computing resource scheduling strategy, the computing resource allocation is adjusted in advance.

[0194] According to some embodiments, a monitoring system based on Prometheus and Grafana can be deployed. Through the Exporter component of Prometheus, the service quality indicator data of each tenant, including response time and success rate, can be collected in real time, and the real-time data can be stored in the database; the ETL tool Apache Nifi can be used to extract the service quality indicator data from the time series database, and the statistical data such as the average response time, the standard deviation of the response time, and the success rate can be calculated. Based on the statistical data, the data analysis tool Pandas can be used to generate a service quality report, including the statistical data and analysis results of each indicator. The analysis results can include a trend chart of the response time, a fluctuation chart of the success rate, etc.

[0195] In step S25, a performance indicator of each computing node in the at least one computing node is obtained;

[0196] In some embodiments, the number corresponding to the performance indicator may be, for example, at least one. The performance indicator does not specifically refer to a fixed indicator. For example, when a modification instruction for the performance indicator is received, the performance indicator may also change accordingly. For example, when the number of indicators corresponding to the performance indicator changes, the number of performance indicators may also change accordingly.

[0197] In some embodiments, the performance indicator may include, for example, at least one of computing resource utilization and response time. The performance indicator may also vary accordingly for different computing resource allocation scenarios.

[0198] According to some embodiments, a performance indicator of each computing node in the at least one computing node is obtained.

[0199] In step S26, a computing resource allocation strategy effect evaluation report corresponding to each computing node is obtained according to the performance indicator;

[0200] In some embodiments, the computing power resource allocation strategy effect evaluation report can be used to indicate, for example, an estimate of the effect after allocation before the computing power resource allocation strategy is used to allocate computing power resources. The computing power resource allocation strategy effect evaluation report does not specifically refer to a fixed report. For example, when the evaluation criteria or evaluation conditions change, the computing power resource allocation strategy effect evaluation report may also change accordingly. For example, when the acquired performance indicators change, the computing power resource allocation strategy effect evaluation report may also change accordingly. Among them, the evaluation criteria or evaluation conditions may be, for example, pre-acquired. The evaluation criteria may, for example, include the percentage of increased resource utilization and the percentage of shortened response time.

[0201] According to some embodiments, the allocation effect of the computing power resource allocation strategy can be judged based on the performance indicators, and a computing power resource allocation strategy effect evaluation report can be generated, wherein the computing power resource allocation strategy effect evaluation report can, for example, describe in detail the degree to which the computing power resource allocation strategy improves various performance indicators, such as improved resource utilization and reduced response time, and can also analyze abnormal situations that may occur during the optimization process and their impacts, such as intensified resource competition, and based on the evaluation results, put forward suggestions and plans for further optimization.

[0202] In step S27, the target computing resource allocation strategy is adjusted according to the computing resource allocation strategy effect evaluation report to obtain an adjusted target computing resource allocation strategy;

[0203] The specific process is as described above and will not be repeated here.

[0204] According to some embodiments, the performance indicators and computing resource scheduling strategies that need to be optimized can be determined based on the computing resource allocation strategy effect evaluation report, and the parameters of the computing resource competition model and the optimization target of the dynamic computing resource scheduling algorithm can be adjusted. According to the optimized computing resource competition model and the dynamic computing resource scheduling algorithm, the optimized computing resource scheduling strategy parameters are obtained, and the optimized computing resource scheduling strategy is implemented.

[0205] In some embodiments, for example, when a target computing power resource allocation strategy is determined to be optimized based on the computing power resource allocation strategy effect evaluation report, the target computing power resource allocation strategy can be adjusted to obtain an adjusted target computing power resource allocation strategy.

[0206] In step S28, the computing resources of each computing node in the at least one computing node are adjusted using the adjusted target computing resource allocation strategy.

[0207] In some embodiments, when the adjusted target computing power resource allocation strategy is obtained, the adjusted target computing power resource allocation strategy can be used to adjust the computing power resources of each computing node in the at least one computing node.

[0208] According to some embodiments, the solutions of the embodiments of the present disclosure may also include, for example, resource over-allocation strategies, edge computing and fog computing, obtaining migration costs when migrating tasks, and using data encryption and privacy protection technologies during the migration process.

[0209] In some embodiments, when optimizing the distributed computing architecture, edge computing and fog computing technologies can be introduced to reduce the burden on the central server, reduce data transmission delays, and improve computing efficiency. At the same time, it supports task migration in a multi-cloud environment, enhances the flexibility and fault tolerance of the system, and avoids dependence on a single cloud service provider.

[0210] In some embodiments, when resource utilization is improved, a resource over-allocation strategy can be used to further improve resource utilization while ensuring system stability. At the same time, a resource recovery mechanism is established to timely recover and reuse idle resources to maximize resource utilization efficiency.

[0211] In some embodiments, when optimizing the migration strategy in multiple dimensions, not only dependencies and resource usage can be considered during task migration, but also the migration cost can be comprehensively evaluated, including data transmission cost, downtime cost, etc., to achieve comprehensive optimization of migration cost and system performance. In addition, according to the service quality assurance agreement (SLA) of different tenants, the migration strategy is customized to ensure that the differentiated needs of each tenant are met.

[0212] In some embodiments, in terms of security and privacy protection, data encryption and privacy protection technologies can be used during task migration to ensure the security and privacy of tenant data. At the same time, a fine-grained access control mechanism is introduced to ensure that only authorized users and tasks can access and use corresponding resources, thereby improving the security of the system.

[0213] In some or related embodiments, performance indicators of each computing node in the at least one computing node are obtained; a computing power resource allocation strategy effectiveness evaluation report corresponding to each computing node is obtained based on the performance indicators; the target computing power resource allocation strategy is adjusted based on the computing power resource allocation strategy effectiveness evaluation report to obtain an adjusted target computing power resource allocation strategy; the computing power resources of each computing node in the at least one computing node are adjusted using the adjusted target computing power resource allocation strategy, and the evaluation report can be obtained through the performance indicators to dynamically adjust the target computing power resource allocation strategy, thereby improving the accuracy of obtaining the target computing power resource allocation strategy and improving the accuracy of computing power resource allocation.

[0214] A block diagram of a computing resource matching device according to an exemplary embodiment. Figure 3 , the device 300 comprises:

[0215] The set acquisition unit 301 is used to identify the state of the acquired computing resource usage data set corresponding to at least one computing node by using the computing resource competition model, and acquire the computing resource competition state data set;

[0216] The set acquisition unit 301 is used to acquire a scheduling suggestion data set according to the computing power resource competition state data set;

[0217] A target determination unit 302 is used to determine a scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set and the scheduling suggestion data set, and obtain a first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and the genetic algorithm;

[0218] The resource matching unit 303 is used to use the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set as the target computing power resource allocation strategy, and use the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node.

[0219] According to some embodiments, the set acquisition unit 301 is further configured to:

[0220] Obtain a collection of historical computing resource usage data through the database;

[0221] Preprocessing the historical computing resource usage data set to obtain a processed historical computing resource usage data set, wherein the preprocessing includes at least one of removing outliers, filling missing values, and data normalization;

[0222] Using an isolation forest algorithm to identify the processed historical computing power resource usage data set, and obtaining at least one abnormal data in the processed historical computing power resource usage data set;

[0223] Annotating the at least one abnormal data to obtain an annotated historical computing resource usage data set;

[0224] The annotated historical computing power resource usage data set and the Bayesian network algorithm are used to perform model training to obtain a computing power resource competition model.

[0225] According to some embodiments, the set acquisition unit 301 is used to determine the scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set and the scheduling recommendation data set, and to acquire the first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and the genetic algorithm, specifically for:

[0226] Determining computing resource utilization and response time as scheduling optimization targets according to the computing resource usage data set, the computing resource competition status data set, and the scheduling suggestion data set;

[0227] According to the scheduling optimization objective and the genetic algorithm, a second computing power resource allocation strategy set corresponding to at least one computing node is obtained;

[0228] A multivariate linear regression model is used to analyze the system performance based on the historical resource utilization and the historical response time, and a first weight corresponding to the computing resource utilization and a second weight corresponding to the response time are obtained;

[0229] Obtaining, according to the computing power resource utilization rate, the response time, the first weight, the second weight, and the fitness function, a fitness value corresponding to each computing power resource allocation strategy in the second computing power resource allocation strategy set;

[0230] Obtain computing power resource allocation strategies whose fitness values ​​are greater than the fitness threshold in the second computing power resource allocation strategy set and add them to the third computing power resource allocation strategy set;

[0231] According to a preset mutation probability, cross-processing is performed on the third computing power resource allocation strategy set to obtain a fourth computing power resource allocation strategy set;

[0232] A genetic algorithm is used to optimize the fourth computing power resource allocation strategy set to obtain the first computing power resource allocation strategy set.

[0233] According to some embodiments, the resource matching unit 303 is used, after matching the computing power resources of each computing node in the at least one computing node by using the target computing power resource allocation strategy, to:

[0234] Obtaining a performance indicator of each computing node in the at least one computing node;

[0235] Obtaining a computing power resource allocation strategy effect evaluation report corresponding to each computing node according to the performance indicator;

[0236] According to the computing power resource allocation strategy effect evaluation report, the target computing power resource allocation strategy is adjusted to obtain an adjusted target computing power resource allocation strategy;

[0237] The computing resources of each computing node in the at least one computing node are adjusted using the adjusted target computing resource allocation strategy.

[0238] According to some embodiments, the resource matching unit 303 is used to match the computing resources of each computing node in the at least one computing node by using the target computing resource allocation strategy, specifically to:

[0239] Acquire monitoring data corresponding to each computing node through a sensor, wherein the monitoring data includes a computing resource usage data set and a service quality indicator data set;

[0240] Using a target sensitivity classification model to identify the task set corresponding to each computing node, and obtaining a sensitivity classification result corresponding to each task in the task set;

[0241] When the monitoring data meets the task migration triggering condition, obtaining the computing node for task migration;

[0242] Determine, according to the current network topology information, the computing resource usage data of each computing node, and the sensitivity classification result, a migration task having a sensitivity greater than a sensitivity threshold;

[0243] The Dijkstra algorithm is used to calculate the shortest path, migration time and used resources from the source computing node to the target computing node, and determine the task migration strategy, wherein the target computing node is a node among the at least one computing node that meets the task migration requirements;

[0244] The task migration strategy is adopted to migrate the migration task from the source computing node to the target computing node.

[0245] According to some embodiments, the resource matching unit 303 is further specifically configured to:

[0246] According to the computing power resource usage data set of each computing node, the computing power resource usage data set is preprocessed to obtain a processed computing power resource usage data set;

[0247] Using a sliding window technology, the computing resource usage data of each tenant in the tenant set within the sliding window is obtained, wherein the tenant set corresponds to each computing node;

[0248] Get the average computing resource usage data in each time window;

[0249] According to the preset usage rate threshold set and the average computing resource usage data, a first time period and a second time period are acquired, wherein the first time period is a time period in which the usage rate is greater than a first preset usage rate threshold in the usage rate threshold set, and the second time period is a time period in which the usage rate is less than a second preset usage rate threshold in the usage rate threshold set, and the first usage rate threshold is greater than the second usage rate threshold;

[0250] Obtaining the computing power resource contention degree of each tenant in the first time period according to the computing power resource contention degree algorithm corresponding to the first time period;

[0251] The task migration strategy is adjusted according to the computing power resource competition degree to obtain an adjusted task migration strategy.

[0252] According to some embodiments, the resource matching unit 303 is used to match the computing resources of each computing node in the at least one computing node by using the target computing resource allocation strategy, specifically to:

[0253] Call data based on the task computing resources corresponding to each computing node;

[0254] Using the Apriori algorithm and the K-means clustering algorithm, the dependency relationship between at least one task corresponding to each computing node is obtained;

[0255] According to the dependency relationship between the tasks and the processing time log of each task, obtaining the processing time of each task;

[0256] According to the processing time of each task, a task dependency data set is obtained, wherein the task dependency data set includes the execution time of the task, the execution time of the dependent tasks and the computing resource consumption index;

[0257] Acquire, according to the task dependency data set, a time sensitivity of a task dependency between the first task and a second task, wherein the first task is any task of at least one task, and the second task is any task of the at least one task except the first task;

[0258] Obtaining a task migration order according to the time sensitivity of the task dependencies;

[0259] The computing resources of each computing node in the at least one computing node are matched according to the target computing resource allocation strategy in the task migration order.

[0260] According to some embodiments, the resource matching unit 303 is used to match the computing resources of each computing node in the at least one computing node by using the target computing resource allocation strategy, specifically to:

[0261] Using a computing power resource demand prediction model and a computing power resource usage data set to obtain computing power resource demand information corresponding to each computing task;

[0262] The computing power resources of each computing node in the at least one computing node are matched according to the computing power resource demand information and the target computing power resource allocation strategy.

[0263] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0264] In some or related embodiments, a set acquisition unit is used to use a computing power resource competition model to perform state identification on a computing power resource usage data set corresponding to at least one computing node, and obtain a computing power resource competition state data set; the set acquisition unit is used to obtain a scheduling recommendation data set based on the computing power resource competition state data set; the target determination unit is used to determine a scheduling optimization target based on the computing power resource usage data set, the computing power resource competition state data set and the scheduling recommendation data set, and obtain a first computing power resource allocation strategy set corresponding to at least one computing node based on the scheduling optimization target and a genetic algorithm; a resource matching unit is used to use the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set as the target computing power resource allocation strategy, and use the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node. Therefore, the computing power resource allocation strategy can be dynamically adjusted through the computing power resource usage data, the matching of the computing power resource allocation strategy with the current allocation scenario can be improved, the computing power resource allocation strategy can be optimized, and the competition for computing power resources that leads to low computing power resource utilization can be reduced. The intelligence of the computing power resource allocation strategy can be improved, the utilization rate of computing power resources can be improved, and the overall performance of the system can be improved.

[0265] Figure 44 is a block diagram of a network device 400 provided in an embodiment of the present disclosure. For example, the network device 400 may be provided as a network device. Figure 4 , the network device 400 includes a processing component 422, which further includes at least one processor, and a memory resource represented by a memory 432 for storing instructions executable by the processing component 422, such as an application. The application stored in the memory 432 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 422 is configured to execute instructions to perform any of the methods described above in the network device.

[0266] The network device 400 may also include a power supply component 427 configured to perform power management of the network device 400, a wired or wireless network interface 450 configured to connect the network device 400 to a network, and an input / output (I / O) interface 458. The network device 400 may operate based on an operating system stored in the memory 432, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM or the like.

[0267] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0268] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0269] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0270] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0271] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0272] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0273] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0274] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A computing resource matching method, characterized in that: include: Using a computing power resource competition model to identify the state of the computing power resource usage data set corresponding to at least one computing node, and obtaining a computing power resource competition state data set; Acquire a scheduling suggestion data set according to the computing power resource competition state data set; Determine a scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set, and the scheduling suggestion data set, and obtain a first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and the genetic algorithm; The computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set is used as the target computing power resource allocation strategy, and the computing power resources of each computing node in the at least one computing node are matched using the target computing power resource allocation strategy.

2. The method according to claim 1, characterized in that: The method further comprises: Obtain a collection of historical computing resource usage data through the database; Preprocessing the historical computing resource usage data set to obtain a processed historical computing resource usage data set, wherein the preprocessing includes at least one of removing outliers, filling missing values, and data normalization; Using an isolation forest algorithm to identify the processed historical computing power resource usage data set, and obtaining at least one abnormal data in the processed historical computing power resource usage data set; Annotating the at least one abnormal data to obtain an annotated historical computing resource usage data set; The annotated historical computing power resource usage data set and the Bayesian network algorithm are used to perform model training to obtain a computing power resource competition model.

3. The method according to claim 1, characterized in that: The determining of a scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set, and the scheduling suggestion data set, and acquiring a first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and a genetic algorithm, includes: Determining computing resource utilization and response time as scheduling optimization targets according to the computing resource usage data set, the computing resource competition status data set, and the scheduling suggestion data set; According to the scheduling optimization objective and the genetic algorithm, a second computing power resource allocation strategy set corresponding to at least one computing node is obtained; A multivariate linear regression model is used to analyze the system performance based on the historical resource utilization and the historical response time, and a first weight corresponding to the computing resource utilization and a second weight corresponding to the response time are obtained; Obtaining, according to the computing power resource utilization rate, the response time, the first weight, the second weight, and the fitness function, a fitness value corresponding to each computing power resource allocation strategy in the second computing power resource allocation strategy set; Obtain computing power resource allocation strategies whose fitness values ​​are greater than the fitness threshold in the second computing power resource allocation strategy set and add them to the third computing power resource allocation strategy set; According to a preset mutation probability, cross-processing is performed on the third computing power resource allocation strategy set to obtain a fourth computing power resource allocation strategy set; A genetic algorithm is used to optimize the fourth computing power resource allocation strategy set to obtain the first computing power resource allocation strategy set.

4. The method according to claim 1, characterized in that: After matching the computing resources of each computing node in the at least one computing node by using the target computing resource allocation strategy, the method further includes: Obtaining a performance indicator of each computing node in the at least one computing node; Obtaining a computing power resource allocation strategy effect evaluation report corresponding to each computing node according to the performance indicator; According to the computing power resource allocation strategy effect evaluation report, the target computing power resource allocation strategy is adjusted to obtain an adjusted target computing power resource allocation strategy; The computing resources of each computing node in the at least one computing node are adjusted using the adjusted target computing resource allocation strategy.

5. The method according to claim 1, characterized in that The adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes: Acquire monitoring data corresponding to each computing node through a sensor, wherein the monitoring data includes a computing resource usage data set and a service quality indicator data set; Using a target sensitivity classification model to identify the task set corresponding to each computing node, and obtaining a sensitivity classification result corresponding to each task in the task set; When the monitoring data meets the task migration triggering condition, obtaining the computing node for task migration; Determine, according to the current network topology information, the computing resource usage data of each computing node, and the sensitivity classification result, a migration task having a sensitivity greater than a sensitivity threshold; The Dijkstra algorithm is used to calculate the shortest path, migration time and used resources from the source computing node to the target computing node, and determine the task migration strategy, wherein the target computing node is a node among the at least one computing node that meets the task migration requirements; The task migration strategy is adopted to migrate the migration task from the source computing node to the target computing node.

6. The method according to claim 5, characterized in that The method further comprises: According to the computing power resource usage data set of each computing node, the computing power resource usage data set is preprocessed to obtain a processed computing power resource usage data set; Using a sliding window technology, the computing resource usage data of each tenant in the tenant set within the sliding window is obtained, wherein the tenant set corresponds to each computing node; Get the average computing resource usage data in each time window; According to the preset usage rate threshold set and the average computing resource usage data, a first time period and a second time period are acquired, wherein the first time period is a time period in which the usage rate is greater than a first preset usage rate threshold in the usage rate threshold set, and the second time period is a time period in which the usage rate is less than a second preset usage rate threshold in the usage rate threshold set, and the first usage rate threshold is greater than the second usage rate threshold; Obtaining the computing power resource contention degree of each tenant in the first time period according to the computing power resource contention degree algorithm corresponding to the first time period; The task migration strategy is adjusted according to the computing power resource competition degree to obtain an adjusted task migration strategy.

7. The method according to claim 1, characterized in that The adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes: Call data based on the task computing resources corresponding to each computing node; Using the Apriori algorithm and the K-means clustering algorithm, the dependency relationship between at least one task corresponding to each computing node is obtained; According to the dependency relationship between the tasks and the processing time log of each task, obtaining the processing time of each task; According to the processing time of each task, a task dependency data set is obtained, wherein the task dependency data set includes the execution time of the task, the execution time of the dependent tasks and the computing resource consumption index; Acquire, according to the task dependency data set, a time sensitivity of a task dependency between the first task and a second task, wherein the first task is any task of at least one task, and the second task is any task of the at least one task except the first task; Obtaining a task migration order according to the time sensitivity of the task dependencies; The computing resources of each computing node in the at least one computing node are matched according to the target computing resource allocation strategy in the task migration order.

8. The method according to claim 1, characterized in that: The adopting the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node includes: Using a computing power resource demand prediction model and a computing power resource usage data set to obtain computing power resource demand information corresponding to each computing task; The computing power resources of each computing node in the at least one computing node are matched according to the computing power resource demand information and the target computing power resource allocation strategy.

9. A computing resource matching device, characterized in that: include: A set acquisition unit, used to identify the state of the acquired computing resource usage data set corresponding to at least one computing node using a computing resource competition model, and acquire a computing resource competition state data set; The set acquisition unit is used to acquire a scheduling suggestion data set according to the computing power resource competition state data set; a target determination unit, configured to determine a scheduling optimization target according to the computing power resource usage data set, the computing power resource competition state data set, and the scheduling suggestion data set, and to obtain a first computing power resource allocation strategy set corresponding to at least one computing node according to the scheduling optimization target and a genetic algorithm; A resource matching unit is used to use the computing power resource allocation strategy with the highest fitness value in the first computing power resource allocation strategy set as the target computing power resource allocation strategy, and use the target computing power resource allocation strategy to match the computing power resources of each computing node in the at least one computing node.

10. A network device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the computing power resource matching method as described in any one of claims 1 to 7.

Images