Load scheduling method and system based on kubernetes

Abstract

The invention provides a kubernetes-based load scheduling method and system. The method comprises the following steps: acquiring a demand annotation of a game process container to be deployed, and GPU computing power, network delay and load indexes of each cluster node in a kubernetes cluster; inputting the GPU computing power, the network delay and the load index of each cluster node into a node state recognition model to obtain a state recognition result of each cluster node; selecting candidate cluster nodes from the cluster nodes according to the demand annotation and the state recognition result of each cluster node; inputting the state recognition result, the demand annotation and the game service scene features of each candidate cluster node into a deep reinforcement learning model to obtain a priority score of each candidate cluster node; and selecting a target cluster node with the highest priority score from the candidate cluster nodes to deploy the game process container. And the load scheduling requirements of high-performance computing scenes such as game services and the like can be met.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a load scheduling method and system based on Kubernetes. Background Art

[0002] With the rapid development of cloud-native technology, Kubernetes has become the de facto standard for enterprise-level containerized application deployment and management due to its container orchestration capabilities, high scalability, and standardized interfaces. However, existing scheduling strategies mostly rely on hard-coded rules to filter nodes, such as filtering by a fixed latency threshold. Such methods do not dynamically evaluate the real-time status of nodes, resulting in scheduling decisions lagging behind actual load requirements, and thus unable to meet the load scheduling requirements in high-performance computing scenarios such as game services. Summary of the Invention

[0003] The technical problem to be solved by this application is to provide a load scheduling method and system based on Kubernetes that can meet the load scheduling requirements in high-performance computing scenarios such as game services. The specific solutions are as follows:

[0004] A load scheduling method based on Kubernetes includes:

[0005] Obtain the requirement annotations of the game process containers to be deployed, the GPU computing power, network latency, and load metrics of each cluster node in the Kubernetes cluster;

[0006] Input the GPU computing power, network latency, and load metrics of each cluster node into a pre-trained node status recognition model to obtain the status recognition results of each cluster node;

[0007] Select candidate cluster nodes from each of the cluster nodes according to the requirement annotations and the status recognition results of each cluster node;

[0008] Input the status recognition results of each candidate cluster node, the requirement annotations, and the current game business scenario characteristics into a deep reinforcement learning model to obtain the priority scores of each candidate cluster node;

[0009] Select the target cluster node with the highest priority score from each of the candidate cluster nodes to deploy the game process container.

[0010] In the above method, optionally, the node status recognition model includes a shared encoder, a first prediction branch network, a second prediction branch network, and a third prediction branch network;

[0011] Inputting the GPU computing power, network latency, and load metrics of each of the cluster nodes into a pre-trained node status recognition model to obtain the status recognition result of each of the cluster nodes, including:

[0012] For each of the cluster nodes, input the GPU computing power, network latency, and load metrics of the cluster node into the shared encoder of the node status recognition model to obtain the status fusion feature of the cluster node; input the status fusion feature into the first prediction branch network, the second prediction branch network, and the third prediction branch network respectively to obtain the GPU status prediction information output by the first prediction branch network, the network status prediction information output by the second prediction branch network, and the load status prediction information output by the third prediction branch network; use the GPU status prediction information, network status prediction information, and load status prediction information as the status recognition result of the cluster node.

[0013] In the above method, optionally, the training process of the node status recognition model includes:

[0014] Obtain an initial model to be trained and a first training dataset, where the first training dataset includes multiple first training sample data and the sample label of each first training sample data;

[0015] Train the initial model using the first training dataset;

[0016] When the initial model meets the first training completion condition, determine the initial model that meets the first training completion condition as the trained node status recognition model.

[0017] In the above method, optionally, selecting candidate cluster nodes from each of the cluster nodes according to the requirement annotation and the status recognition result of each of the cluster nodes includes:

[0018] Generate a first constraint condition according to the hardware performance threshold and network latency threshold in the requirement annotation;

[0019] Determine the nodes that meet the first constraint condition among each of the cluster nodes as the initial cluster nodes;

[0020] Filter each of the initial cluster nodes according to the status recognition result of each of the initial cluster nodes and the business scenario type indicated by the requirement annotation to obtain an optimized node set;

[0021] Perform priority sorting on the optimized node set according to a preset node sorting strategy, and intercept a node subset that meets the preset quantity constraint as the candidate cluster node set. The candidate cluster node set includes multiple candidate cluster nodes, and the sorting strategy is dynamically adjusted according to the matching degree between the resource supply ability of each node in the optimized node set and the container resource requirements.

[0022] For the above method, optionally, after selecting the target cluster node with the highest priority score among the candidate cluster nodes and deploying the game process container, the following steps are further included:

[0023] Monitor the runtime metric data of the game process container on the target node, and generate a monitoring data set including the actual resource utilization rate, network communication quality, and load change rate;

[0024] By comparing the monitoring data set with the predicted data of the status recognition result, calculate the node status prediction deviation index, and adjust the node label update frequency of the target cluster node according to the node status prediction deviation index;

[0025] When the load growth rate in the monitoring data set continuously exceeds the corresponding predicted value and reaches a preset threshold, mark the target cluster node as a predicted failure state.

[0026] A load scheduling system based on Kubernetes, including:

[0027] An acquisition unit, configured to acquire the requirement annotation of the game process container to be deployed, the GPU computing power, network latency, and load metrics of each cluster node in the Kubernetes cluster;

[0028] A first execution unit, configured to input the GPU computing power, network latency, and load metrics of each cluster node into a pre-trained node status recognition model to obtain the status recognition result of each cluster node;

[0029] A first selection unit, configured to select candidate cluster nodes from each cluster node according to the requirement annotation and the status recognition result of each cluster node;

[0030] A second execution unit, configured to input the status recognition result of each candidate cluster node, the requirement annotation, and the current game business scenario characteristics into a deep reinforcement learning model to obtain the priority score of each candidate cluster node;

[0031] A third execution unit, configured to select the target cluster node with the highest priority score among the candidate cluster nodes and deploy the game process container.

[0032] For the above system, optionally, the node status recognition model includes a shared encoder, a first prediction branch network, a second prediction branch network, and a third prediction branch network;

[0033] The first execution unit includes:

[0034] The first execution subunit is configured to, for each of the cluster nodes, input the GPU computing power, network latency, and load metrics of the cluster node into the shared encoder of the node status recognition model to obtain the status fusion feature of the cluster node; input the status fusion feature into the first prediction branch network, the second prediction branch network, and the third prediction branch network respectively to obtain the GPU status prediction information output by the first prediction branch network, the network status prediction information output by the second prediction branch network, and the load status prediction information output by the third prediction branch network; and use the GPU status prediction information, network status prediction information, and load status prediction information as the status recognition result of the cluster node.

[0035] For the above system, optionally, the first execution unit includes:

[0036] The first acquisition subunit is configured to acquire an initial model to be trained and a first training dataset, where the first training dataset includes a plurality of first training sample data and sample labels of each first training sample data;

[0037] The training subunit is configured to train the initial model using the first training dataset;

[0038] The second execution subunit is configured to, when the initial model meets the first training completion condition, determine the initial model that meets the first training completion condition as the trained node status recognition model.

[0039] For the above system, optionally, the first selection unit includes:

[0040] The generation subunit is configured to generate a first constraint condition according to the hardware performance threshold and network latency threshold in the requirement annotation;

[0041] The determination subunit is configured to determine the nodes that meet the first constraint condition among the respective cluster nodes as the initial cluster nodes;

[0042] The third execution subunit is configured to screen the respective initial cluster nodes according to the status recognition results of the respective initial cluster nodes and the service scenario type indicated by the requirement annotation to obtain an optimized node set;

[0043] A sorting subunit, configured to perform priority sorting on the optimized node set according to a preset node sorting strategy, and intercept a node subset that meets a preset quantity constraint as a candidate cluster node set, where the candidate cluster node set includes multiple candidate cluster nodes, and the sorting strategy is dynamically adjusted according to the matching degree between the resource supply capacity of each node in the optimized node set and the container resource requirements.

[0044] Optionally, the above system further includes:

[0045] A monitoring unit, configured to monitor the runtime metric data of the game process container on the target node, and generate a monitoring data set including the actual resource utilization rate, network communication quality, and load change rate;

[0046] A calculation unit, configured to calculate a node state prediction deviation index by comparing the monitoring data set with the predicted data of the state recognition result, and adjust the node label update frequency of the target cluster node according to the node state prediction deviation index;

[0047] A fourth execution unit, configured to mark the target cluster node as a predicted failure state when the load growth rate in the monitoring data set continuously exceeds the corresponding predicted value and reaches a preset threshold.

[0048] Based on the load scheduling method and system based on kubernetes provided in the above embodiments of the present application, where the demand annotation of the game process container to be deployed, the GPU computing power, network latency, and load metrics of each cluster node in the kubernetes cluster are obtained; the GPU computing power, network latency, and load metrics of each cluster node are input into a pre-trained node state recognition model to obtain the state recognition result of each cluster node; according to the demand annotation and the state recognition result of each cluster node, candidate cluster nodes are selected from each of the cluster nodes; the state recognition result, the demand annotation, and the current game business scenario characteristics of each candidate cluster node are input into a deep reinforcement learning model to obtain the priority score of each candidate cluster node; and the target cluster node with the highest priority score is selected from each of the candidate cluster nodes to deploy the game process container. Applying the method provided in the embodiments of the present application can meet the load scheduling requirements of high-performance computing scenarios such as game services. Description of the Drawings

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0050] Figure 1 The method flow chart of a load scheduling method based on kubernetes provided for this application;

[0051] Figure 2 The flow chart of the training process of a node status recognition model provided for this application;

[0052] Figure 3 The flow chart of the process of selecting candidate cluster nodes from each cluster node provided for this application;

[0053] Figure 4 The structural schematic diagram of a load scheduling system based on kubernetes provided for this application. Detailed implementation manners

[0054] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.

[0055] In this application, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, the element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

[0056] An embodiment of the present invention provides a load scheduling method based on kubernetes, which is applied to an electronic device. The method flow chart of the method is as Figure 1 shown, and specifically includes:

[0057] S101: Obtain the requirement annotation of the game process container to be deployed, the GPU computing power, network latency, and load metrics of each cluster node in the kubernetes cluster.

[0058] S102: Input the GPU computing power, network latency, and load metrics of each said cluster node into a pre-trained node status recognition model to obtain the status recognition result of each said cluster node.

[0059] S103: Select candidate cluster nodes from each of the cluster nodes according to the requirement annotation and the status recognition result of each cluster node.

[0060] S104: Input the status recognition result of each candidate cluster node, the requirement annotation, and the current game service scenario features into the deep reinforcement learning model to obtain the priority score of each candidate cluster node.

[0061] S105: Select the target cluster node with the highest priority score from each of the candidate cluster nodes to deploy the game process container.

[0062] Applying the method provided in the embodiments of the present application can meet the load scheduling requirements of high-performance computing scenarios such as game services.

[0063] In an embodiment provided by the present application, based on the above solution, optionally, the node status recognition model includes a shared encoder, a first prediction branch network, a second prediction branch network, and a third prediction branch network;

[0064] The step of inputting the GPU computing power, network latency, and load metrics of each cluster node into a pre-trained node status recognition model to obtain the status recognition result of each cluster node includes:

[0065] For each cluster node, input the GPU computing power, network latency, and load metrics of the cluster node into the shared encoder of the node status recognition model to obtain the status fusion feature of the cluster node; input the status fusion feature into the first prediction branch network, the second prediction branch network, and the third prediction branch network respectively to obtain the GPU status prediction information output by the first prediction branch network, the network status prediction information output by the second prediction branch network, and the load status prediction information output by the third prediction branch network; use the GPU status prediction information, network status prediction information, and load status prediction information as the status recognition result of the cluster node.

[0066] In an embodiment provided by the present application, based on the above solution, optionally, the training process of the node status recognition model is as Figure 2 shown and includes:

[0067] S201: Obtain an initial model to be trained and a first training data set. The first training data set includes multiple first training sample data and the sample label of each first training sample data;

[0068] S202: Train the initial model using the first training data set;

[0069] S203: When the initial model meets the first training completion condition, determine the initial model that meets the first training completion condition as the trained node status recognition model.

[0070] In an embodiment provided by the present application, based on the above solution, optionally, the process of selecting candidate cluster nodes from each of the cluster nodes according to the requirement annotation and the status recognition result of each cluster node is as Figure 3 shown and includes:

[0071] S301: Generate a first constraint condition according to the hardware performance threshold and network latency threshold in the requirement annotation.

[0072] S302: Determine the nodes that meet the first constraint condition in each of the cluster nodes as the initial cluster nodes.

[0073] S303: Screen each of the initial cluster nodes according to the status recognition result of each initial cluster node and the business scenario type indicated by the requirement annotation to obtain an optimized node set.

[0074] S304: Perform priority sorting on the optimized node set according to a preset node sorting strategy, and intercept a node subset that meets the preset quantity constraint as the candidate cluster node set. The candidate cluster node set includes multiple candidate cluster nodes, and the sorting strategy is dynamically adjusted according to the matching degree between the resource supply capacity of each node in the optimized node set and the container resource demand.

[0075] In an embodiment provided by the present application, based on the above solution, optionally, after deploying the game process container on the target cluster node with the highest priority score selected from each of the candidate cluster nodes, it further includes:

[0076] Monitor the runtime metric data of the game process container on the target node to generate a monitoring data set including the actual resource utilization rate, network communication quality, and load change rate;

[0077] Calculate the node status prediction deviation index by comparing the monitoring data set with the predicted data of the status recognition result, and adjust the node label update frequency of the target cluster node according to the node status prediction deviation index;

[0078] When the load growth rate in the monitoring data set continuously exceeds the corresponding predicted value and reaches a preset threshold, mark the target cluster node as a predicted failure state.

[0079] See Figure 4, which is a schematic structural diagram of a load scheduling method system based on kubernetes provided by an embodiment of the present application. The system includes:

[0080] An acquisition unit 401, configured to acquire requirement annotations of game process containers to be deployed, GPU computing power, network latency, and load metrics of each cluster node in a kubernetes cluster;

[0081] A first execution unit 402, configured to input the GPU computing power, network latency, and load metrics of each cluster node into a pre-trained node status recognition model to obtain a status recognition result of each cluster node;

[0082] A first selection unit 403, configured to select candidate cluster nodes from each cluster node according to the requirement annotations and the status recognition results of each cluster node;

[0083] A second execution unit 404, configured to input the status recognition results of each candidate cluster node, the requirement annotations, and the current game business scenario characteristics into a deep reinforcement learning model to obtain a priority score of each candidate cluster node;

[0084] A third execution unit 405, configured to select a target cluster node with the highest priority score from each candidate cluster node to deploy game process containers.

[0085] In an embodiment provided by the present application, based on the above solution, optionally, the node status recognition model includes a shared encoder, a first prediction branch network, a second prediction branch network, and a third prediction branch network;

[0086] The first execution unit includes:

[0087] A first execution subunit, configured to, for each cluster node, input the GPU computing power, network latency, and load metrics of the cluster node into the shared encoder of the node status recognition model to obtain a status fusion feature of the cluster node; input the status fusion feature into the first prediction branch network, the second prediction branch network, and the third prediction branch network respectively to obtain GPU status prediction information output by the first prediction branch network, network status prediction information output by the second prediction branch network, and load status prediction information output by the third prediction branch network; use the GPU status prediction information, network status prediction information, and load status prediction information as the status recognition result of the cluster node.

[0088] In an embodiment provided by the present application, based on the above solution, optionally, the first execution unit includes:

[0089] A first acquisition subunit, configured to acquire an initial model to be trained and a first training data set, where the first training data set includes a plurality of first training sample data and sample labels of each first training sample data;

[0090] A training subunit, configured to train the initial model by using the first training data set;

[0091] A second execution subunit, configured to, when the initial model meets a first training completion condition, determine the initial model that meets the first training completion condition as a trained node status recognition model.

[0092] In an embodiment provided by the present application, based on the above solution, optionally, the first selection unit includes:

[0093] A generation subunit, configured to generate a first constraint condition according to a hardware performance threshold and a network delay threshold in the requirement annotation;

[0094] A determination subunit, configured to determine nodes in each of the cluster nodes that meet the first constraint condition as initial cluster nodes;

[0095] A third execution subunit, configured to screen each of the initial cluster nodes according to a status recognition result of each of the initial cluster nodes and a service scenario type indicated by the requirement annotation, to obtain an optimized node set;

[0096] A sorting subunit, configured to perform priority sorting on the optimized node set according to a preset node sorting strategy, and intercept a node subset that meets a preset quantity constraint as a candidate cluster node set, where the candidate cluster node set includes a plurality of candidate cluster nodes, and the sorting strategy is dynamically adjusted according to a matching degree between a resource supply capability and a container resource requirement of each node in the optimized node set.

[0097] In an embodiment provided by the present application, based on the above solution, optionally, the system further includes:

[0098] A monitoring unit, configured to monitor runtime metric data of the game process container on a target node, and generate a monitoring data set including an actual resource utilization rate, network communication quality, and load change rate;

[0099] A calculation unit, configured to calculate a node status prediction deviation index by comparing the monitoring data set with predicted data of the status recognition result, and adjust a node label update frequency of the target cluster node according to the node status prediction deviation index;

[0100] A fourth execution unit is configured to mark the target cluster node as a predicted failure state when the load growth rate in the monitoring data set continuously exceeds the corresponding predicted value and reaches a preset threshold.

[0101] It should be noted that the various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0102] Finally, it should also be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0103] For the convenience of description, when describing the above system, it is divided into various units according to functions for separate description. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0104] From the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present application.

[0105] The above has introduced in detail a load scheduling method based on kubernetes provided by the present application. Specific examples are used in this text to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A load scheduling method based on Kubernetes, characterized in that Including: Obtain the requirement annotation of the game process container to be deployed, the GPU computing power, network latency, and load metrics of each cluster node in the kubernetes cluster; Input the GPU computing power, network latency, and load metrics of each cluster node into a pre-trained node status recognition model to obtain the status recognition result of each cluster node; Select candidate cluster nodes from each of the cluster nodes according to the requirement annotation and the status recognition result of each cluster node; Input the status recognition result of each candidate cluster node, the requirement annotation, and the current game business scenario characteristics into a deep reinforcement learning model to obtain the priority score of each candidate cluster node; Select the target cluster node with the highest priority score from each of the candidate cluster nodes to deploy the game process container.

2. The method according to claim 1, characterized in that, The node status recognition model includes a shared encoder, a first prediction branch network, a second prediction branch network, and a third prediction branch network; The step of inputting the GPU computing power, network latency, and load metrics of each cluster node into a pre-trained node status recognition model to obtain the status recognition result of each cluster node includes: For each cluster node, input the GPU computing power, network latency, and load metrics of the cluster node into the shared encoder of the node status recognition model to obtain the status fusion feature of the cluster node; input the status fusion feature into the first prediction branch network, the second prediction branch network, and the third prediction branch network respectively to obtain the GPU status prediction information output by the first prediction branch network, the network status prediction information output by the second prediction branch network, and the load status prediction information output by the third prediction branch network; use the GPU status prediction information, network status prediction information, and load status prediction information as the status recognition result of the cluster node.

3. The method according to claim 2, wherein The training process of the node status recognition model includes: Obtain an initial model to be trained and a first training data set. The first training data set includes a plurality of first training sample data and the sample label of each first training sample data; the first training sample data includes historical GPU computing power, historical network latency, and historical load metrics; Train the initial model using the first training data set; When the initial model meets the first training completion condition, determine the initial model that meets the first training completion condition as the trained node status recognition model.

4. The method according to claim 1, wherein The step of selecting candidate cluster nodes from each of the cluster nodes according to the requirement annotation and the status recognition result of each cluster node includes: Generate a first constraint condition according to the hardware performance threshold and network latency threshold in the requirement annotation; Determine the nodes that meet the first constraint condition among each of the cluster nodes as the initial cluster nodes; Filter each of the initial cluster nodes according to the status recognition result of each initial cluster node and the business scenario type indicated by the requirement annotation to obtain an optimized node set; Prioritize the optimized node set according to a preset node sorting strategy, and intercept a node subset that meets the preset quantity constraint as the candidate cluster node set. The candidate cluster node set includes multiple candidate cluster nodes, and the sorting strategy is dynamically adjusted according to the matching degree between the resource supply capacity of each node in the optimized node set and the container resource demand.

5. The method according to claim 1, characterized in that, After selecting the target cluster node with the highest priority score among the candidate cluster nodes to deploy the game process container, it further includes: Monitor the runtime metric data of the game process container on the target node, and generate a monitoring data set including the actual resource utilization rate, network communication quality, and load change rate; By comparing the monitoring data set with the predicted data of the status recognition result, calculate the node status prediction deviation index, and adjust the node label update frequency of the target cluster node according to the node status prediction deviation index; When the load growth rate in the monitoring data set continuously exceeds the corresponding predicted value and reaches a preset threshold, mark the target cluster node as a predicted failure state.

6. A load scheduling system based on Kubernetes, characterized in that, It includes: An acquisition unit for acquiring the requirement annotation of the game process container to be deployed, the GPU computing power, network latency, and load metrics of each cluster node in the kubernetes cluster; A first execution unit for inputting the GPU computing power, network latency, and load metrics of each cluster node into a pre-trained node status recognition model to obtain the status recognition result of each cluster node; A first selection unit for selecting candidate cluster nodes from each cluster node according to the requirement annotation and the status recognition result of each cluster node; A second execution unit for inputting the status recognition result of each candidate cluster node, the requirement annotation, and the current game business scenario characteristics into a deep reinforcement learning model to obtain the priority score of each candidate cluster node; A third execution unit for selecting the target cluster node with the highest priority score among the candidate cluster nodes to deploy the game process container.

7. The system according to claim 6, wherein The node status recognition model includes a shared encoder, a first prediction branch network, a second prediction branch network, and a third prediction branch network; The first execution unit includes: A first execution subunit for inputting the GPU computing power, network latency, and load metrics of each cluster node into the shared encoder of the node status recognition model for each cluster node to obtain the status fusion feature of the cluster node; Input the status fusion feature into the first prediction branch network, the second prediction branch network, and the third prediction branch network respectively to obtain the GPU status prediction information output by the first prediction branch network, the network status prediction information output by the second prediction branch network, and the load status prediction information output by the third prediction branch network; use the GPU status prediction information, network status prediction information, and load status prediction information as the status recognition result of the cluster node.

8. The system according to claim 7, wherein The first execution unit includes: A first acquisition subunit, configured to acquire an initial model to be trained and a first training data set, where the first training data set includes a plurality of first training sample data and sample labels of each first training sample data; A training subunit, configured to train the initial model by using the first training data set; A second execution subunit, configured to, when the initial model meets a first training completion condition, determine the initial model that meets the first training completion condition as a trained node status recognition model.

9. The system according to claim 6, wherein The first selection unit includes: A generation subunit, configured to generate a first constraint condition according to the hardware performance threshold and the network latency threshold in the requirement annotation; A determination subunit, configured to determine the nodes that meet the first constraint condition among the respective cluster nodes as initial cluster nodes; A third execution subunit, configured to screen the respective initial cluster nodes according to the status recognition results of the respective initial cluster nodes and the business scenario type indicated by the requirement annotation to obtain an optimized node set; A sorting subunit, configured to perform priority sorting on the optimized node set according to a preset node sorting strategy, and intercept a node subset that meets a preset quantity constraint as a candidate cluster node set, where the candidate cluster node set includes a plurality of candidate cluster nodes, and the sorting strategy is dynamically adjusted according to the matching degree between the resource supply capacity of each node in the optimized node set and the container resource requirement.

10. The system according to claim 6, characterized in that, It further includes: A monitoring unit, configured to monitor the runtime metric data of the game process container on a target node and generate a monitoring data set including an actual resource utilization rate, network communication quality, and load change rate; A calculation unit, configured to calculate a node status prediction deviation index by comparing the monitoring data set with the predicted data of the status recognition result, and adjust the node label update frequency of the target cluster node according to the node status prediction deviation index; A fourth execution unit, configured to mark the target cluster node as a predicted failure state when the load growth rate in the monitoring data set continuously exceeds the corresponding predicted value and reaches a preset threshold.

Images