Performance scheduling method and system based on AI cloud server

Abstract

The present invention belongs to the technical field of cloud server management, and specifically is a performance scheduling method and system based on AI cloud servers, wherein the system comprises a data acquisition module, an AI analysis module, a performance scheduling module, a monitoring feedback module and a background management terminal; the present invention collects relevant data of the cloud server through the data acquisition module, the AI ​​analysis module uses deep learning technology to perform real-time analysis on the received data, and predicts the load trend and user demand in the future period of time, the performance scheduling module dynamically adjusts the resource allocation of the cloud server according to the prediction result of the AI ​​analysis module, and the monitoring feedback module monitors the performance status of the cloud server in real time to optimize the performance scheduling strategy for the cloud server, thereby improving the performance and service quality of the cloud server, and by analyzing the scheduling rate performance and scheduling effect performance of the cloud server, it is conducive to timely reminding management personnel to make corresponding improvement measures and improve the operating performance of the cloud server.

Description

Technical Field

[0001] The present invention relates to the field of cloud server management technology, and in particular to a performance scheduling method and system based on an AI cloud server. Background Art

[0002] Cloud servers are an important part of cloud computing services. With the rapid development of cloud computing technology, cloud servers have become an important platform for enterprises and individual users to process data, deploy applications and provide services. The performance scheduling of cloud servers is a key factor affecting the quality of cloud services.

[0003] Traditional cloud server performance scheduling methods are often based on static rules and preset parameters, resulting in low resource utilization and unstable service quality. They are also unable to reasonably analyze and accurately evaluate the performance scheduling efficiency and performance scheduling effect for cloud servers, making it difficult for managers to make corresponding improvement measures in a timely manner, and it is difficult to effectively improve the operating performance of cloud servers. The degree of intelligence is low.

[0004] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a performance scheduling method and system based on AI cloud server, which solves the problems of low resource utilization and unstable service quality caused by the prior art, and the inability to reasonably analyze and accurately evaluate the performance scheduling efficiency and performance scheduling effect for the cloud server, making it difficult to effectively improve the operating performance of the cloud server and having a low degree of intelligence.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The performance scheduling system based on AI cloud server includes data collection module, AI analysis module, performance scheduling module, monitoring feedback module and background management terminal;

[0008] The data collection module collects relevant data of the cloud servers to be monitored, including real-time load data, user request data, and application operation data, and sends the collected data to the AI ​​analysis module;

[0009] The AI ​​analysis module uses deep learning technology to perform real-time analysis on the received data, predict the load trend and user demand in the future, and send the prediction results to the performance scheduling module; the performance scheduling module dynamically adjusts the resource allocation of the cloud server based on the prediction results of the AI ​​analysis module;

[0010] The monitoring feedback module monitors the performance status of the cloud server in real time and feeds back the performance data to the AI ​​analysis module and the backend management end to optimize the performance scheduling strategy for the cloud server.

[0011] Furthermore, the backend management terminal communicates with the scheduling rapid inspection and evaluation module, which is used to analyze the scheduling rate performance of the cloud server during the detection period, generate a scheduling speed difference signal or a scheduling speed optimization signal through analysis, and send the scheduling speed difference signal to the backend management terminal. When the backend management terminal receives the scheduling speed difference signal, it will issue a corresponding warning.

[0012] Furthermore, the specific analysis process of the dispatching rapid inspection evaluation module is as follows:

[0013] When the performance scheduling module dynamically adjusts the resource allocation of the cloud server, the duration of the corresponding adjustment process is collected and marked as the scheduling duration, and the ratio of the scheduling duration to the corresponding preset scheduling duration threshold is marked as the scheduling condition value;

[0014] All dispatching condition values ​​within the detection period are obtained and their average is used to calculate the dispatching time measurement value, and the number of dispatching time condition values ​​that exceed the preset dispatching time condition threshold within the detection period is marked as the dispatching excess value; the dispatching quick evaluation value is obtained by numerically calculating the dispatching time measurement value and the dispatching excess value, and the dispatching quick evaluation value is numerically compared with the preset dispatching quick evaluation threshold. If the dispatching quick evaluation value exceeds the preset dispatching quick evaluation threshold, a dispatching quick abnormal signal is generated; if the dispatching quick evaluation value does not exceed the preset dispatching quick evaluation threshold, a dispatching quick excellent signal is generated.

[0015] Furthermore, the scheduling quick inspection and evaluation module is communicatively connected to the scheduling effectiveness inspection and evaluation module. The scheduling quick inspection and evaluation module sends the scheduling speed optimization signal to the scheduling effectiveness inspection and evaluation module. When the scheduling effectiveness inspection and evaluation module receives the scheduling speed optimization signal, it analyzes the scheduling effect of the cloud server during the detection period, generates a scheduling effectiveness error signal or a scheduling effectiveness optimization signal through analysis, and sends the scheduling effectiveness error signal to the background management end. When the background management end receives the scheduling effectiveness error signal, it issues a corresponding warning.

[0016] Furthermore, the specific analysis process of the dispatching efficiency inspection and evaluation module is as follows:

[0017] The start time of the corresponding adjustment process and the end time of the previous group of adjustment processes performed by the performance scheduling module are collected, and the interval between the start time of the corresponding adjustment process and the end time of the previous group of adjustment processes is marked as the interval time;

[0018] The inter-adjustment time is numerically compared with the preset inter-adjustment time threshold. If the inter-adjustment time does not exceed the preset inter-adjustment time threshold, the corresponding inter-adjustment time is marked as the out-of-tune time; if the inter-adjustment time exceeds the preset inter-adjustment time threshold, the corresponding inter-adjustment time is marked as the optimal adjustment time; the number of out-of-tune times in the detection period is obtained and marked as the out-of-tune detection value, and the out-of-tune detection value is numerically compared with the preset out-of-tune detection threshold. If the out-of-tune detection value exceeds the preset out-of-tune detection threshold, a scheduling effect abnormality signal is generated.

[0019] Furthermore, if the out-of-tune detection value does not exceed the preset out-of-tune detection threshold, all inter-tune durations are sorted in order from the front to the back, the number of out-of-tune durations between two adjacent groups of optimal tuning durations is marked as an out-of-tune value, the out-of-tune value is numerically compared with the preset out-of-tune value threshold, and the number of out-of-tune values ​​exceeding the preset out-of-tune value threshold during the detection period is marked as an over-tune detection value, and the out-of-tune value with the largest value is marked as an out-of-tune amplitude value;

[0020] The server inspection value is obtained by performing operation detection, evaluation and analysis on the cloud server, and the scheduling efficiency evaluation value is obtained by numerically calculating the server inspection value, anomaly detection value, over-holding number detection value and anomaly number amplitude. The scheduling efficiency evaluation value is numerically compared with the preset scheduling efficiency evaluation threshold. If the scheduling efficiency evaluation value exceeds the preset scheduling efficiency evaluation threshold, a scheduling efficiency abnormality signal is generated; if the scheduling efficiency evaluation value does not exceed the preset scheduling efficiency evaluation threshold, a scheduling efficiency excellent signal is generated.

[0021] Furthermore, the analysis and acquisition method of the server evaluation value is as follows:

[0022] Several detection time periods are set during the detection period, and the response data, throughput data, and error rate data of the cloud server in the corresponding detection time period are collected. The time period analysis value is obtained by numerically calculating the response data, throughput data, and error rate data, and the time period analysis value is numerically compared with the preset time period analysis threshold. If the time period analysis value exceeds the preset time period analysis threshold, the corresponding detection time period is marked as a server abnormal evaluation time period;

[0023] The number of server anomaly time periods during the detection period is obtained and marked as server anomaly values, and the server anomaly value is obtained by averaging all time period analysis values ​​during the detection period, and the time period analysis value with the largest value is marked as the time period anomaly amplitude value; the server inspection value is obtained by numerically calculating the server anomaly value, the server anomaly analysis value and the time period anomaly amplitude value.

[0024] Furthermore, the backend management end is communicated with the auxiliary supervision module, which is used to analyze the operating risks of the cloud server, generate a risk warning signal or a risk qualification signal through analysis, and send the risk warning signal to the backend management end. When the backend management end receives the risk warning signal, it will issue a corresponding warning.

[0025] Furthermore, the specific operation process of the auxiliary supervision module includes:

[0026] Based on the security log of the cloud server, all security risk events of the cloud server during the detection period are obtained, all security risk events are classified, the number of security risk events of the corresponding type is collected and marked as the risk number detection value, each type of security risk event is set in advance to correspond to a set of preset risk impact values, and the product of the risk number detection value of the corresponding type of security risk event and the corresponding preset risk impact value is marked as the risk detection value;

[0027] Obtain the hidden danger detection values ​​of all types of safety hidden danger events that occurred during the detection period, and mark the sum of all hidden danger detection values ​​as the hidden danger comprehensive analysis value, compare the hidden danger comprehensive analysis value with the preset hidden danger comprehensive analysis threshold, and generate a risk warning signal if the hidden danger comprehensive analysis value exceeds the preset hidden danger comprehensive analysis threshold;

[0028] If the hidden danger analysis value does not exceed the preset hidden danger analysis threshold, the impact time of the corresponding security hidden danger event on the cloud server is compared with the preset impact time threshold of the corresponding type of security hidden danger event. If the impact time exceeds the preset impact time threshold, the corresponding security hidden danger event is marked as a damage event; the number of damage events in the detection period is obtained and marked as the damage detection value, and the ratio of the impact time of the corresponding security hidden danger event to the corresponding preset impact time threshold is marked as the impact time value, and the impact time values ​​of all security hidden danger events in the detection period are averaged to obtain the impact time measurement value;

[0029] The server risk analysis value is obtained by numerically calculating the hidden danger analysis value, damage detection value and impact time measurement value, and the server risk analysis value is numerically compared with the preset server risk analysis threshold. If the server risk analysis value exceeds the preset server risk analysis threshold, a risk warning signal is generated; if the server risk analysis value does not exceed the preset server risk analysis threshold, a risk qualification signal is generated.

[0030] Furthermore, the present invention also proposes a performance scheduling method based on an AI cloud server, comprising the following steps:

[0031] Step 1: Collect relevant data of the cloud server to be monitored;

[0032] Step 2: Use deep learning technology to analyze the received data in real time to predict the load trend and user demand in the future;

[0033] Step 3: Dynamically adjust the resource allocation of the cloud server according to the prediction results;

[0034] Step 4: Monitor the performance status of the cloud server in real time and optimize the performance scheduling strategy for the cloud server accordingly.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. In the present invention, the data collection module collects relevant data of the cloud server, the AI ​​analysis module uses deep learning technology to perform real-time analysis on the received data, and predicts the load trend and user demand in the future. The performance scheduling module dynamically adjusts the resource allocation of the cloud server according to the prediction results of the AI ​​analysis module, and the monitoring feedback module monitors the performance status of the cloud server in real time to optimize the performance scheduling strategy for the cloud server, thereby improving the performance and service quality of the cloud server. In addition, by analyzing the scheduling rate performance and scheduling effect performance of the cloud server, it is helpful to timely remind the management personnel to make corresponding improvement measures and improve the operating performance of the cloud server;

[0037] 2. In the present invention, the operating risk of the cloud server is analyzed through the auxiliary supervision module, and a risk warning signal or a risk qualification signal is generated through the analysis. When the risk warning signal is generated, the background management end issues a warning, which can reasonably analyze and accurately evaluate the degree of safety hazards of the cloud server during the detection period, so as to promptly remind the management personnel to strengthen the security supervision of the cloud server, ensure the safety and stability of the cloud server, and further improve the operating performance of the cloud server. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0039] Figure 1 This is a system block diagram of Embodiment 1 of the present invention;

[0040] Figure 2 This is a system block diagram of Embodiment 2 of the present invention;

[0041] Figure 3 This is a flow chart of the method of Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Embodiment 1: Figure 1 As shown, the performance scheduling system based on the AI ​​cloud server proposed in the present invention includes a data acquisition module, an AI analysis module, a performance scheduling module, a monitoring feedback module and a background management terminal;

[0044] The data acquisition module collects relevant data of the cloud servers to be monitored, including real-time load data (referring to the load status of the cloud server at a certain moment, including key performance indicators such as CPU usage, memory occupancy, disk I / O, etc. The real-time load data can directly reflect the current operating status of the cloud server and is an important indicator for evaluating whether the server performance meets business needs), user request data (including various request information sent by users, such as HTTP requests, database query requests, etc. User request data can reflect the user's demand and usage of the server, and is an important basis for evaluating server performance and user experience) and application operation data (referring to the operation status information of various applications running on the cloud server, such as process status, resource occupancy, etc. Application operation data can directly reflect the operation efficiency and performance of the application on the server, and is a key indicator for evaluating application quality and service quality), etc., and sends the collected data to the AI ​​analysis module.

[0045] The AI ​​analysis module uses deep learning technology (deep learning model) to perform real-time analysis on the received data, predict the load trend and user demand in the future, and send the prediction results to the performance scheduling module; and after receiving the raw data, the AI ​​analysis module pre-processes it, including data cleaning, denoising, missing value filling and other steps to ensure the accuracy and consistency of the data.

[0046] The performance scheduling module dynamically adjusts the resource allocation of the cloud server based on the prediction results of the AI ​​analysis module to ensure that resource utilization is maximized while meeting user needs, thereby improving server performance and service quality; the details are as follows:

[0047] After receiving the prediction results, further analysis and interpretation of the prediction results will take into account various factors, such as resource competition between different applications, the overall health of the server, the availability of network bandwidth, and the user's quality of service (QoS) requirements;

[0048] Based on the analysis of the forecast data, formulate a set of appropriate resource scheduling strategies, including increasing or decreasing the CPU, memory or storage resources of a specific application, reallocating network bandwidth, or starting / shutting down additional virtual machine instances. The goal of the scheduling strategy is to maximize the overall performance of the server while meeting user needs and ensure efficient use of resources.

[0049] Once the scheduling strategy is formulated, the performance scheduling module performs actual resource scheduling operations through the cloud server's management interface (such as API), including starting / stopping virtual machines, adjusting virtual machine configurations, and reallocating network resources.

[0050] The monitoring feedback module monitors the performance status of the cloud server in real time, and feeds back the performance data to the AI ​​analysis module and the background management end to optimize the performance scheduling strategy for the cloud server (that is, if it is found that the scheduling strategy fails to achieve the expected effect, such as the server load is too high, the response time is prolonged, etc., the scheduling strategy will be adjusted in time and the resource scheduling operation will be re-executed. The present invention uses AI technology to analyze server load and user needs in real time, dynamically adjust resource allocation, improve the performance and service quality of the cloud server, and provide users with a better experience.

[0051] In addition, the backend management terminal is connected to the scheduling speed inspection and evaluation module, which analyzes the scheduling rate performance of the cloud server during the detection period. Preferably, the detection period is fifteen days; generates a scheduling speed difference signal or a scheduling speed optimization signal through analysis, and sends the scheduling speed difference signal to the backend management terminal;

[0052] When the backend management receives the scheduling speed abnormality signal, it issues a corresponding warning, which can reasonably analyze and accurately evaluate the scheduling efficiency performance of the cloud server, thereby reminding the management personnel to take corresponding improvement measures in time to ensure the subsequent scheduling efficiency, thereby improving the operating performance of the cloud server; the specific analysis process of the scheduling speed inspection evaluation module is as follows:

[0053] When the performance scheduling module dynamically adjusts the resource allocation of the cloud server, the duration of the corresponding adjustment process is collected and marked as the scheduling duration, wherein the larger the value of the scheduling duration, the slower the scheduling efficiency of the corresponding process; the ratio of the scheduling duration to the corresponding preset scheduling duration threshold is marked as the scheduling condition value; all scheduling condition values ​​within the detection period are obtained and averaged to calculate the scheduling time measurement value, and the number of scheduling condition values ​​that exceed the preset scheduling condition threshold within the detection period is marked as the scheduling excess value;

[0054] The scheduling quick evaluation value LX is obtained by numerically calculating the scheduling measurement value LK and the scheduling excess value LP through the formula LX=mu1*LK+mu2*LP / mu1, wherein mu1 and mu2 are preset proportional coefficients, and mu1>mu2>0; and the larger the value of the scheduling quick evaluation value LX, the worse the scheduling efficiency performance of the cloud server during the detection period is overall;

[0055] The scheduling quick evaluation value LX is numerically compared with the preset scheduling quick evaluation threshold. If the scheduling quick evaluation value LX exceeds the preset scheduling quick evaluation threshold, it indicates that the scheduling efficiency performance for the cloud server during the detection period is generally poor, and a scheduling quick abnormal signal is generated; if the scheduling quick evaluation value LX does not exceed the preset scheduling quick evaluation threshold, it indicates that the scheduling efficiency performance for the cloud server during the detection period is generally good, and a scheduling quick excellent signal is generated.

[0056] Further, the scheduling speed inspection and evaluation module is connected to the scheduling efficiency inspection and evaluation module in communication, and the scheduling speed inspection and evaluation module sends the scheduling speed optimization signal to the scheduling efficiency inspection and evaluation module. When the scheduling efficiency inspection and evaluation module receives the scheduling speed optimization signal, it analyzes the scheduling effect of the cloud server during the detection period, generates a scheduling efficiency error signal or a scheduling efficiency optimization signal through analysis, and sends the scheduling efficiency error signal to the background management end;

[0057] When the backend management receives the scheduling effect abnormality signal, it issues a corresponding warning, which can reasonably analyze and accurately evaluate the scheduling effect performance of the cloud server, thereby reminding the management personnel to take corresponding improvement measures in time to ensure the subsequent scheduling effect and further improve the operating performance of the cloud server; the specific analysis process of the scheduling effect inspection and evaluation module is as follows:

[0058] The start time of the corresponding adjustment process and the end time of the previous group of adjustment processes of the performance scheduling module are collected, and the interval between the start time of the corresponding adjustment process and the end time of the previous group of adjustment processes is marked as the adjustment interval; the smaller the value of the adjustment interval is, the shorter the interval between the two groups of adjustment processes is, and the greater the probability that the scheduling effect of the corresponding adjustment process is poor;

[0059] Compare the inter-adjustment time with the preset inter-adjustment time threshold. If the inter-adjustment time does not exceed the preset inter-adjustment time threshold, it indicates that the interval between the corresponding two groups of adjustment processes is short, and the corresponding inter-adjustment time is marked as the out-of-adjustment time; if the inter-adjustment time exceeds the preset inter-adjustment time threshold, the corresponding inter-adjustment time is marked as the optimal adjustment time; obtain the number of out-of-adjustment times during the detection period and mark it as the out-of-adjustment detection value, compare the out-of-adjustment detection value with the preset out-of-adjustment detection threshold, and if the out-of-adjustment detection value exceeds the preset out-of-adjustment detection threshold, it indicates that the scheduling effect of the cloud server during the detection period is poor, and a scheduling effect abnormality signal is generated;

[0060] If the out-of-tune detection value does not exceed the preset out-of-tune detection threshold, all inter-tune durations are sorted in order from the front to the back, the number of out-of-tune durations between two adjacent groups of optimal tuning durations is marked as an out-of-tune value, the out-of-tune value is numerically compared with the preset out-of-tune threshold, and the number of out-of-tune values ​​exceeding the preset out-of-tune threshold during the detection period is marked as an over-tune detection value, and the out-of-tune value with the largest value is marked as an out-of-tune amplitude;

[0061] The server evaluation value is obtained by performing operation detection, evaluation and analysis on the cloud server, specifically: setting a number of detection periods within the detection period, and collecting the response data, throughput data and error rate data of the cloud server within the corresponding detection period;

[0062] Among them, the response data refers to the time required for the cloud server to process user requests. A shorter response time indicates that the performance scheduling is effective and can quickly meet user needs. The throughput data refers to the number of requests that the cloud server can process per unit time. High throughput means that the cloud server can efficiently process a large number of concurrent requests. The error rate data refers to the error ratio of the server when processing requests. A low error rate indicates good scheduling performance.

[0063] By formula The response data XR, the throughput data XD and the error rate data XP are numerically calculated to obtain the period analysis value XN, wherein up1, up2 and up3 are preset proportional coefficients whose values ​​are greater than zero, and the larger the value of the period analysis value XN is, the worse the operation performance of the cloud server in the corresponding detection period is, and the worse the scheduling effect of the cloud server is;

[0064] Compare the period analysis value XN with the preset period analysis threshold. If the period analysis value XN exceeds the preset period analysis threshold, it indicates that the operation performance of the cloud server in the corresponding detection period is poor and the scheduling effect of the cloud server is poor. Then, the corresponding detection period is marked as a server abnormal evaluation period.

[0065] The number of server abnormality time periods during the detection period is obtained and marked as server abnormality values, and the server abnormality values ​​are obtained by averaging all time period analysis values ​​during the detection period, and the time period analysis value with the largest value is marked as the time period abnormality amplitude value;

[0066] The server abnormality value YG, the server abnormality value YW and the period abnormality value YF are numerically calculated by the formula YP=e1*YG+(e2*YW+e3*YF) / 2 to obtain the server evaluation value YP; wherein, e1, e2, and e3 are preset proportional coefficients, e1>e2>0>0; and the larger the value of the server evaluation value YP, the worse the overall operation performance of the cloud server during the detection period;

[0067] The server inspection value YP, the out-of-tune detection value WN, the over-holding value WK and the out-of-tune amplitude WP are numerically calculated by the formula WY=sg1*WS+(sg2*WN+sg3*WK+sg4*WP) / sg1 to obtain the scheduling effectiveness evaluation value WY; wherein, sg1, sg2, sg3, sg4 are preset proportional coefficients with values ​​greater than zero, and the larger the value of the scheduling effectiveness evaluation value WY, the worse the scheduling effect performance for the cloud server during the detection period;

[0068] The scheduling efficiency evaluation value WY is numerically compared with the preset scheduling efficiency evaluation threshold. If the scheduling efficiency evaluation value WY exceeds the preset scheduling efficiency evaluation threshold, it indicates that the scheduling effect performance for the cloud server during the detection period is generally poor, and a scheduling efficiency error signal is generated; if the scheduling efficiency evaluation value WY does not exceed the preset scheduling efficiency evaluation threshold, it indicates that the scheduling effect performance for the cloud server during the detection period is generally good, and a scheduling efficiency improvement signal is generated.

[0069] Embodiment 2: Figure 2 As shown, the difference between this embodiment and embodiment 1 is that the backend management end is connected to the auxiliary supervision module in communication, and the auxiliary supervision module is used to analyze the operation risk of the cloud server, generate a risk warning signal or a risk qualified signal through analysis, and send the risk warning signal to the backend management end;

[0070] When the backend management receives a risk warning signal, it issues a corresponding warning, which can reasonably analyze and accurately evaluate the degree of security risks of the cloud server during the detection period, so as to promptly remind the management personnel to strengthen the security supervision of the cloud server, such as enhancing or continuously optimizing the security mechanism of the cloud server, ensuring the security and stability of the cloud server, thereby further improving the operating performance of the cloud server; the specific operation process of the auxiliary supervision module is as follows:

[0071] Based on the security log of the cloud server, all security risk events (including network attacks, virus intrusions, etc.) of the cloud server during the detection period are obtained, all security risk events are classified, the number of security risk events of the corresponding type is collected and marked as the risk number detection value, and each type of security risk event is set in advance to correspond to a set of preset risk impact values, wherein the values ​​of the preset risk impact values ​​are all positive numbers, and the greater the risk brought by the corresponding type of security risk event to the server, the greater the value of the preset risk impact value adapted to it; the product of the risk number detection value of the corresponding type of security risk event and the corresponding preset risk impact value is marked as the risk detection value;

[0072] Obtain the hidden danger detection values ​​of all types of security hidden danger events that occurred during the detection period, and mark the sum of all hidden danger detection values ​​as the hidden danger comprehensive analysis value, wherein the smaller the hidden danger comprehensive analysis value is, the smaller the security hidden danger suffered by the cloud server during the detection period is; compare the hidden danger comprehensive analysis value with the preset hidden danger comprehensive analysis threshold value, if the hidden danger comprehensive analysis value exceeds the preset hidden danger comprehensive analysis threshold value, it indicates that the security hidden danger suffered by the cloud server during the detection period is relatively large, and then generate a risk warning signal;

[0073] If the risk analysis value does not exceed the preset risk analysis threshold, the impact duration of the corresponding security risk event on the cloud server (such as the duration of the corresponding network attack) is compared with the preset impact duration threshold of the corresponding type of security risk event. If the impact duration exceeds the preset impact duration threshold, it indicates that the corresponding security risk event has a significant adverse impact on the cloud server, and the corresponding security risk event is marked as a damage event.

[0074] The number of damage-causing events during the detection period is obtained and marked as the damage detection value, and the ratio of the impact duration of the corresponding potential safety hazard event to the corresponding preset impact duration threshold is marked as the impact time value, and the impact time values ​​of all potential safety hazard events during the detection period are averaged to obtain the impact time measurement value;

[0075] The server risk analysis value FY is obtained by numerically calculating the hidden danger comprehensive analysis value FL, the damage detection value FK and the impact time measurement value FN through the formula FY=(b1*FL+b2*FK) / 2+b3*FN, wherein b1, b2 and b3 are preset proportional coefficients, b3>b2>b1>0; and the larger the value of the server risk analysis value FY is, the greater the overall operation risk of the cloud server during the detection period is;

[0076] The server risk analysis value FY is numerically compared with the preset server risk analysis threshold. If the server risk analysis value FY exceeds the preset server risk analysis threshold, it indicates that the overall operation risk of the cloud server during the detection period is relatively large, and a risk warning signal is generated; if the server risk analysis value FY does not exceed the preset server risk analysis threshold, it indicates that the overall operation risk of the cloud server during the detection period is relatively small, and a risk qualification signal is generated.

[0077] Embodiment 3: Figure 3 As shown, the difference between this embodiment and embodiment 1 and embodiment 2 is that the performance scheduling method based on the AI ​​cloud server proposed in the present invention includes the following steps:

[0078] Step 1: Collect relevant data of the cloud server to be monitored;

[0079] Step 2: Use deep learning technology to analyze the received data in real time to predict the load trend and user demand in the future;

[0080] Step 3: Dynamically adjust the resource allocation of the cloud server according to the prediction results;

[0081] Step 4: Monitor the performance status of the cloud server in real time and optimize the performance scheduling strategy for the cloud server accordingly.

[0082] The working principle of the present invention is as follows: when in use, the relevant data of the cloud server to be monitored is collected through the data acquisition module, the AI ​​analysis module uses deep learning technology to perform real-time analysis on the received data to predict the load trend and user demand in the future, the performance scheduling module dynamically adjusts the resource allocation of the cloud server according to the prediction results of the AI ​​analysis module, and the monitoring feedback module monitors the performance status of the cloud server in real time to optimize the performance scheduling strategy for the cloud server, thereby improving the performance and service quality of the cloud server, which is conducive to providing users with a better experience, and the scheduling rate performance of the cloud server during the detection period is analyzed through the scheduling speed inspection and evaluation module to generate a scheduling speed difference signal or a scheduling speed optimization signal. When the scheduling speed optimization signal is generated, the scheduling effect of the cloud server during the detection period is analyzed through the scheduling effect inspection and evaluation module, and a scheduling effect difference signal or a scheduling effect optimization signal is generated through analysis. When the scheduling speed difference signal or the scheduling effect difference signal is generated, the management personnel are reminded to take corresponding improvement measures, thereby further improving the operating performance of the cloud server, and the degree of intelligence is high.

[0083] The above formulas are all dimensionless and numerical calculations. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formula are set by technicians in this field according to actual conditions. The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the contents of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that technicians in the relevant technical field can understand and use the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. The performance scheduling system based on AI cloud server is characterized by: It includes data collection module, AI analysis module, performance scheduling module, monitoring feedback module and background management terminal; The data collection module collects relevant data of the cloud server to be monitored, including real-time load data, user request data, and application running data, and sends the collected data to the AI ​​analysis module; real-time load data refers to the load status of the cloud server at a certain moment, including CPU usage, memory occupancy, and disk I / O. Real-time load data directly reflects the current operating status of the cloud server and is an important indicator for evaluating whether the server performance meets business needs; User request data refers to various request information sent by users, including HTTP requests and database query requests. User request data reflects users' demand for and usage of the server, and is an important basis for evaluating server performance and user experience. Application running data refers to the running status information of various applications running on the cloud server, including process status and resource usage. Application running data directly reflects the running efficiency and performance of the application on the server, and is a key indicator for evaluating application quality and service quality. The AI ​​analysis module uses deep learning technology to perform real-time analysis on the received data, predict the load trend and user demand in the future, and send the prediction results to the performance scheduling module. In addition, after receiving the raw data, the AI ​​analysis module pre-processes it, including data cleaning, denoising, and missing value filling, to ensure the accuracy and consistency of the data. The performance scheduling module dynamically adjusts the resource allocation of the cloud server based on the prediction results of the AI ​​analysis module to ensure that resource utilization is maximized while meeting user needs and improving server performance and service quality; the details are as follows: After receiving the prediction results, further analysis and interpretation of the prediction results will take into account a variety of factors, including resource competition between different applications, the overall health of the server, the availability of network bandwidth, and the user's service quality requirements; Based on the analysis of the forecast data, formulate a set of appropriate resource scheduling strategies, including increasing or decreasing the CPU, memory or storage resources of a specific application, reallocating network bandwidth, or starting / shutting down additional virtual machine instances; the goal of the scheduling strategy is to maximize the overall performance of the server while meeting user needs and ensure efficient use of resources; Once the scheduling strategy is formulated, the performance scheduling module performs the actual resource scheduling operations through the cloud server's management interface, including starting / stopping virtual machines, adjusting virtual machine configurations, and reallocating network resources; The monitoring feedback module monitors the performance status of the cloud server in real time and feeds back the performance data to the AI ​​analysis module and the backend management end to optimize the performance scheduling strategy for the cloud server; The backend management terminal is connected to the auxiliary supervision module, which is used to analyze the operational risk of the cloud server, generate a risk warning signal or a risk qualified signal through analysis, and send the risk warning signal to the backend management terminal. When the backend management terminal receives the risk warning signal, it issues a corresponding warning. The specific operation process of the auxiliary supervision module includes: Based on the security log of the cloud server, all security risk events of the cloud server during the detection period are obtained, all security risk events are classified, the number of security risk events of the corresponding type is collected and marked as the risk number detection value, each type of security risk event is set in advance to correspond to a set of preset risk impact values, and the product of the risk number detection value of the corresponding type of security risk event and the corresponding preset risk impact value is marked as the risk detection value; Obtain the hidden danger detection values ​​of all types of safety hidden danger events that occurred during the detection period, and mark the sum of all hidden danger detection values ​​as the hidden danger comprehensive analysis value. If the hidden danger comprehensive analysis value exceeds the preset hidden danger comprehensive analysis threshold, a risk warning signal is generated; If the hidden danger analysis value does not exceed the preset hidden danger analysis threshold, the impact time of the corresponding security hidden danger event on the cloud server is compared with the preset impact time threshold of the corresponding type of security hidden danger event. If the impact time exceeds the preset impact time threshold, the corresponding security hidden danger event is marked as a damage event; the number of damage events in the detection period is obtained and marked as the damage detection value, and the ratio of the impact time of the corresponding security hidden danger event to the corresponding preset impact time threshold is marked as the impact time value, and the impact time values ​​of all security hidden danger events in the detection period are averaged to obtain the impact time measurement value; The server risk analysis value is obtained by numerically calculating the hidden danger analysis value, damage detection value and impact time measurement value. If the server risk analysis value exceeds the preset server risk analysis threshold, a risk warning signal is generated; if the server risk analysis value does not exceed the preset server risk analysis threshold, a risk qualification signal is generated.

2. The performance scheduling system based on AI cloud server according to claim 1 is characterized in that: The backend management terminal communicates with the scheduling rapid inspection and evaluation module, which is used to analyze the scheduling rate performance of the cloud server during the detection period, generate a scheduling speed difference signal or a scheduling speed optimization signal through analysis, and send the scheduling speed difference signal to the backend management terminal. When the backend management terminal receives the scheduling speed difference signal, it will issue a corresponding warning.

3. The performance scheduling system based on AI cloud server according to claim 2 is characterized in that: The specific analysis process of the dispatch rapid inspection evaluation module is as follows: When the performance scheduling module dynamically adjusts the resource allocation of the cloud server, the duration of the corresponding adjustment process is collected and marked as the scheduling duration, and the ratio of the scheduling duration to the corresponding preset scheduling duration threshold is marked as the scheduling condition value; All dispatching condition values ​​within the detection period are obtained and their average is used to calculate the dispatching time measurement value, and the number of dispatching time condition values ​​that exceed the preset dispatching time condition threshold within the detection period is marked as the dispatching excess value; the dispatching quick evaluation value is obtained by numerically calculating the dispatching time measurement value and the dispatching excess value, and the dispatching quick evaluation value is numerically compared with the preset dispatching quick evaluation threshold. If the dispatching quick evaluation value exceeds the preset dispatching quick evaluation threshold, a dispatching quick abnormal signal is generated; if the dispatching quick evaluation value does not exceed the preset dispatching quick evaluation threshold, a dispatching quick excellent signal is generated.

4. The performance scheduling system based on AI cloud server according to claim 2 is characterized in that: The scheduling quick inspection and evaluation module is communicatively connected to the scheduling effectiveness inspection and evaluation module. The scheduling quick inspection and evaluation module sends the scheduling speed optimization signal to the scheduling effectiveness inspection and evaluation module. When the scheduling effectiveness inspection and evaluation module receives the scheduling speed optimization signal, it analyzes the scheduling effect of the cloud server during the detection period, generates a scheduling effectiveness error signal or a scheduling effectiveness optimization signal through analysis, and sends the scheduling effectiveness error signal to the background management end. When the background management end receives the scheduling effectiveness error signal, it issues a corresponding warning.

5. The performance scheduling system based on AI cloud server according to claim 4 is characterized in that: The specific analysis process of the dispatching effectiveness inspection and evaluation module is as follows: The start time of the corresponding adjustment process and the end time of the previous group of adjustment processes performed by the performance scheduling module are collected, and the interval between the start time of the corresponding adjustment process and the end time of the previous group of adjustment processes is marked as the interval time; The inter-adjustment time is numerically compared with the preset inter-adjustment time threshold. If the inter-adjustment time does not exceed the preset inter-adjustment time threshold, the corresponding inter-adjustment time is marked as the out-of-tune time; if the inter-adjustment time exceeds the preset inter-adjustment time threshold, the corresponding inter-adjustment time is marked as the optimal adjustment time; the number of out-of-tune times in the detection period is obtained and marked as the out-of-tune detection value, and the out-of-tune detection value is numerically compared with the preset out-of-tune detection threshold. If the out-of-tune detection value exceeds the preset out-of-tune detection threshold, a scheduling effect abnormality signal is generated.

6. The performance scheduling system based on AI cloud server according to claim 5 is characterized in that: If the out-of-tune detection value does not exceed the preset out-of-tune detection threshold, all inter-tune durations are sorted in order from the front to the back, the number of out-of-tune durations between two adjacent groups of optimal tuning durations is marked as an out-of-tune value, the out-of-tune value is numerically compared with the preset out-of-tune threshold, and the number of out-of-tune values ​​exceeding the preset out-of-tune threshold during the detection period is marked as an over-tune detection value, and the out-of-tune value with the largest value is marked as an out-of-tune amplitude; The server inspection value is obtained by performing operation detection and evaluation analysis on the cloud server, and the dispatching effectiveness evaluation value is obtained by numerically calculating the server inspection value, the abnormal adjustment detection value, the over-holding number detection value and the abnormal adjustment number amplitude. The dispatching effectiveness evaluation value is numerically compared with the preset dispatching effectiveness evaluation threshold. If the dispatching effectiveness evaluation value exceeds the preset dispatching effectiveness evaluation threshold, a dispatching effectiveness abnormality signal is generated; If the dispatching efficiency evaluation value does not exceed the preset dispatching efficiency evaluation threshold, a dispatching efficiency signal is generated.

7. The performance scheduling system based on AI cloud server according to claim 6 is characterized in that: The specific method for analyzing and obtaining the server evaluation value is as follows: Several detection time periods are set during the detection period, and the response data, throughput data, and error rate data of the cloud server in the corresponding detection time period are collected. The time period analysis value is obtained by numerically calculating the response data, throughput data, and error rate data, and the time period analysis value is numerically compared with the preset time period analysis threshold. If the time period analysis value exceeds the preset time period analysis threshold, the corresponding detection time period is marked as a server abnormal evaluation time period; The number of server abnormality time periods during the detection period is obtained and marked as server abnormality values, and the server abnormality values ​​are obtained by averaging all time period analysis values ​​during the detection period, and the time period analysis value with the largest value is marked as the time period abnormality amplitude value; The server evaluation value is obtained by numerically calculating the server abnormality value, server abnormal analysis value and time period abnormal amplitude value.

8. A performance scheduling method based on AI cloud server, characterized in that: The method adopts the performance scheduling system based on AI cloud server as described in any one of claims 1-7.

Images