A Cloud Information Management Method for an Online Education Platform

Abstract

The present invention relates to the technical field of cloud information management, and particularly to a cloud information management method for an online education platform. The method includes: obtaining the frequency and load data of a refresh request, as well as the CPU utilization rate and bandwidth utilization rate; determining high-frequency refresh nodes according to the fluctuations of the CPU utilization rate and bandwidth utilization rate; screening to obtain high-frequency refresh moments according to the changes in the frequency distribution and load data of the refresh requests under the high-frequency refresh nodes; further determining refresh segments with dense refresh distributions, and determining the unconscious refresh performance evaluation value of the high-frequency refresh nodes according to the load changes at all high-frequency refresh moments within the refresh segments; determining unconscious impact nodes according to the unconscious refresh performance evaluation value of each server node, and restricting the maximum refresh frequency of the unconscious impact nodes. The present invention can accurately identify unconscious refreshes, liberate the cache resources in this part, reduce the system burden, and improve the overall resource utilization efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud information management, and particularly relates to a cloud information management method for an online education platform. Background Art

[0002] With the rapid development of online education, cloud information management has become a core component of the infrastructure of education platforms. Online education platforms need to process a large amount of information such as user data, course content, learning records, and examination results, and at the same time ensure the high availability, scalability, and data security of the platforms. Cloud technologies provide flexible computing resources and storage spaces, enabling the platforms to dynamically adjust their processing capabilities to meet different traffic demands, especially during critical moments such as peak examination periods and live course sessions.

[0003] In related technologies, online education platforms may encounter performance bottlenecks during peak periods such as live classes, for example, problems such as video stuttering, slow page loading, and request timeouts. There will be situations where people repeatedly refresh and click on the page, and the request volume at this node is too large. The system will use load balancing to distribute requests to other nodes; however, sometimes there will be unconscious frequent refreshes by people without encountering performance bottlenecks, which will consume additional bandwidth and computing resources, thus occupying server resources, reducing the response speed of other user requests, increasing the system burden, and affecting the overall resource utilization efficiency. Summary of the Invention

[0004] In order to solve the technical problem in related technologies that unconscious frequent refreshes by people without encountering performance bottlenecks will consume additional bandwidth and computing resources, thus occupying server resources, reducing the response speed of other user requests, increasing the system burden, and affecting the overall resource utilization efficiency, the present invention provides a cloud information management method for an online education platform. The specific technical solutions adopted are as follows:

[0005] The present invention proposes a cloud information management method for an online education platform. The method includes:

[0006] Obtain the frequency and load data of user refresh requests at each collection moment for each server node of the online education platform, as well as multi-dimensional real-time data collected at all collection moments for all server nodes. Among them, the real-time data includes: CPU utilization rate and bandwidth utilization rate;

[0007] According to the fluctuation discreteness and range of the CPU utilization rate and bandwidth utilization rate, determine the CPU peak fluctuation characteristic index and bandwidth peak fluctuation characteristic index for each server node; combine the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index to screen out high-frequency refresh nodes from the server nodes;

[0008] Based on the frequency distribution change and load data change of user refresh requests at each sampling moment of high-frequency refresh nodes, high-frequency refresh moments are screened from the sampling moments of each high-frequency refresh node; according to the temporal distribution of high-frequency refresh moments, refresh segments with dense refresh distributions are determined, and based on the load changes of all high-frequency refresh moments within the refresh segments, the unconscious refresh performance evaluation value of high-frequency refresh nodes is determined;

[0009] Based on the unconscious refresh performance evaluation value of each server node, unconscious influence nodes are determined, and the maximum refresh frequency of unconscious influence nodes is restricted.

[0010] Furthermore, determining the CPU peak fluctuation characteristic index and bandwidth peak fluctuation characteristic index of each server node according to the fluctuation discreteness and range of CPU utilization rate and bandwidth utilization rate includes:

[0011] Calculate the product of the variance and range of CPU utilization rate at all sampling moments of each server node, and perform normalization processing as the CPU peak fluctuation characteristic index;

[0012] Calculate the product of the variance and range of bandwidth utilization rate at all sampling moments of each server node, and perform normalization processing as the bandwidth peak fluctuation characteristic index.

[0013] Furthermore, screening high-frequency refresh nodes from server nodes by combining the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index includes:

[0014] Calculate the product value of the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index, and perform normalization processing as the high-frequency refresh characteristic value;

[0015] Server nodes with the high-frequency refresh characteristic value greater than the preset refresh characteristic threshold are used as high-frequency refresh nodes.

[0016] Furthermore, screening high-frequency refresh moments from the sampling moments of each high-frequency refresh node according to the frequency distribution change and load data change of user refresh requests at each sampling moment of high-frequency refresh nodes includes:

[0017] According to the frequency distribution and discreteness characteristics of user refresh requests at all sampling moments of the same high-frequency refresh node, determine the demand factor of the refresh request at each sampling moment;

[0018] By analogy with the demand factor of the refresh request, calculate the demand factor of the load data at each sampling moment of the same high-frequency refresh node;

[0019] Determine the refresh request volume - load dependency index of the refresh request and the load at each sampling moment according to the correlation and numerical difference between the demand factor of the refresh request and the demand factor of the load data at each sampling moment;

[0020] Take the sampling moments when the refresh request volume - load dependency index is greater than the preset dependency threshold as high - frequency refresh moments.

[0021] Further, the determination of the demand factor of the refresh request at each sampling moment according to the frequency distribution and discreteness characteristics of the user refresh requests at all sampling moments of the same high - frequency refresh node includes:

[0022] Take any sampling moment as the target moment, calculate the mean value of the frequencies of the user refresh requests at all sampling moments of the same high - frequency refresh node, and normalize the difference between the frequency of the user refresh request at the target moment and this mean value as the refresh frequency difference;

[0023] Calculate the negative value of the variance of the frequencies of the user refresh requests at all sampling moments of the same high - frequency refresh node, and perform maximum - minimum normalization to obtain the refresh frequency stability index;

[0024] Take the ratio of the refresh frequency difference and the refresh frequency stability index as the demand factor of the refresh request at the target moment.

[0025] Further, the determination of the refresh request volume - load dependency index of the refresh request and the load at each sampling moment according to the correlation and numerical difference between the demand factor of the refresh request and the demand factor of the load data at each sampling moment includes:

[0026] Calculate the correlation index between the demand factor of the refresh request and the demand factor of the load data at each moment based on the Pearson correlation coefficient;

[0027] Calculate the absolute value of the difference between the demand factor of the refresh request and the demand factor of the load data at any sampling moment, and normalize the negative value of the absolute value of the difference as the demand similarity index corresponding to the sampling moment;

[0028] Calculate the product of the demand similarity index and the correlation index, and perform normalization to obtain the refresh request volume - load dependency index corresponding to the sampling moment.

[0029] Further, the determination of the refresh segment with dense refresh distribution according to the distribution of high - frequency refresh moments in time sequence includes:

[0030] Take the time interval between any high - frequency refresh moment and the previous high - frequency refresh moment as the high - frequency interval corresponding to the high - frequency refresh moment;

[0031] Take the other high-frequency intervals on both sides of any high-frequency interval in terms of time sequence as adjacent intervals, calculate the average duration of the adjacent intervals as the influence interval value of the middle high-frequency interval; calculate the ratio of the duration of the middle high-frequency interval to the influence interval value as the special ratio.

[0032] When the special ratio is greater than the preset ratio, take the middle high-frequency interval as the special interval; take the time interval between the two closest special intervals as the refresh segment with dense refresh distribution.

[0033] Further, determining the unconscious refresh performance evaluation value of the high-frequency refresh node according to the load changes at all high-frequency refresh moments within the refresh segment includes:

[0034] Calculate the average value of the load data at all high-frequency refresh moments within the refresh segment, and perform normalization processing to obtain the load coefficient of the refresh segment.

[0035] Take the refresh segment with the load coefficient less than the preset load threshold as the unconscious segment.

[0036] Determine the unconscious refresh performance evaluation value according to the number of high-frequency refresh moments corresponding to the unconscious segment of the high-frequency refresh node.

[0037] Further, determining the unconscious refresh performance evaluation value according to the number of high-frequency refresh moments corresponding to the unconscious segment of the high-frequency refresh node includes:

[0038] Take the ratio of the number of high-frequency refresh moments corresponding to the unconscious segment of the high-frequency refresh node to the number of all high-frequency refresh moments as the unconscious refresh performance evaluation value.

[0039] Further, determining the unconscious influence node according to the unconscious refresh performance evaluation value of each server node includes:

[0040] Take the server node with the unconscious refresh performance evaluation value greater than the preset performance threshold as the unconscious influence node.

[0041] The present invention has the following beneficial effects:

[0042] The present invention screens high-frequency refresh nodes from server nodes by refreshing the frequency of requests, load data, CPU utilization, and bandwidth utilization. Taking CPU utilization and bandwidth utilization as a group, by specifically analyzing the changes in CPU status and bandwidth status, high-frequency refresh nodes are obtained. Then, for high-frequency refresh nodes, by analyzing the changes in the frequency distribution of refresh requests and load data, the high-frequency refresh moments are determined, that is, the high-frequency refresh moments of the unconscious refresh and stuttering refresh sets are initially screened out. Furthermore, according to the temporal distribution and load changes of the high-frequency refresh moments, the evaluation value of the unconscious refresh performance of high-frequency refresh nodes is determined. The evaluation value of the unconscious refresh performance can accurately represent the unconscious influence nodes that generate unconscious refreshes relatively frequently. By restricting the maximum refresh frequency of the unconscious influence nodes, the cache resources of this part can be liberated, the system burden can be reduced, and the overall resource utilization efficiency can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention 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 some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0044] Figure 1 Flowchart of a cloud information management method for an online education platform provided by an embodiment of the present invention;

[0045] Figure 2 Schematic diagram of CPU utilization curve provided by an embodiment of the present invention;

[0046] Figure 3 Schematic diagram of refresh segmentation provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following, in combination with the drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of a cloud information management method for an online education platform proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0049] The following specifically describes the specific solution of a cloud information management method for an online education platform provided by the present invention in conjunction with the accompanying drawings.

[0050] Please refer to Figure 1 , which shows a flowchart of a cloud information management method for an online education platform provided by an embodiment of the present invention. The method includes:

[0051] S101: Obtain the frequency and load data of user refresh requests at each collection moment of each server node of the online education platform, and multi-dimensional real-time data collected at all collection moments of all server nodes. Among them, the real-time data includes: CPU utilization rate, bandwidth utilization rate.

[0052] With the rapid development of online education, cloud information management has become a core part of the education platform infrastructure. Online education platforms need to process massive amounts of user data, course content, learning records, exam scores, etc., while also ensuring the high availability, scalability, and data security of the platform. Cloud technology provides flexible computing resources and storage space, enabling the platform to dynamically adjust its processing capabilities to meet different traffic demands, especially during critical moments such as exam peaks and live courses.

[0053] Online education platforms may encounter performance bottlenecks during peak periods such as live classes, for example, video stuttering, slow page loading, request timeouts, etc. There will be people repeatedly refreshing and clicking on the page, and the request volume of this node will be too large. The system will distribute requests to other nodes through load balancing; however, sometimes there will be unconscious frequent refreshing without encountering performance bottlenecks, which will consume additional bandwidth and computing resources, thus occupying server resources, reducing the response speed of other user requests, increasing the system burden, and affecting the overall resource utilization efficiency. Therefore, the embodiments of the present invention aim to identify unconscious frequent refreshing and limit it.

[0054] The distributed cloud computing architecture disperses various components of the education platform (such as user management, course management, video live broadcast, exam evaluation, learning analysis, etc.) on servers in multiple geographical locations and collaborates through network connections.

[0055] Therefore, in the embodiments of the present invention, the frequency and load data of user refresh requests at each collection moment of each server node of the online education platform can be obtained. Then, one second can be used as a collection moment, and corresponding multi-dimensional real-time data can be monitored once every 1 second from each component of the online education platform to obtain multi-dimensional real-time data collected at all moments of all server nodes within a day. The real-time data includes: CPU utilization rate, bandwidth utilization rate. The number of user refresh requests and the load situation at each moment of each server node within a day can be obtained by logging at the server side every 1 second as a collection moment.

[0056] S102: Determine the CPU peak fluctuation characteristic index and bandwidth peak fluctuation characteristic index of each server node according to the fluctuation discreteness and range of CPU utilization rate and bandwidth utilization rate; combine the CPU peak fluctuation characteristic index and bandwidth peak fluctuation characteristic index, and screen out high-frequency refresh nodes from the server nodes.

[0057] During peak periods such as live classes, the online education platform may encounter performance bottlenecks, such as video stuttering, slow page loading, request timeouts, etc. Personnel will frequently refresh and click on the page, and frequent requests will cause the server to require more computing resources to process each request. Especially during complex calculations or data processing (such as video decoding, user authentication, data access and storage), the CPU load will increase. If the refresh request involves a large amount of data processing (for example, each refresh requires reloading a complex page), the CPU utilization rate will increase significantly.

[0058] During the process of data monitoring, for the CPU utilization rate of each server node, with time as the abscissa and CPU utilization rate as the ordinate, the CPU utilization rate curve of each node can be obtained. During the non-peak period, the CPU utilization rate curve fluctuates relatively smoothly. When the online education platform encounters a peak period, the CPU utilization rate will increase instantaneously. Therefore, the CPU utilization rate curve of the online education platform during peak periods such as live classes will show large fluctuations, as Figure 2 shown. Figure 2 It is a schematic diagram of the CPU utilization rate curve provided by an embodiment of the present invention.

[0059] At the same time, each page refresh will consume bandwidth. Especially when resources such as video streams and pictures are large, frequent page refreshing will exacerbate the use of network bandwidth, may lead to bandwidth saturation, and further affect the experience of other users. The increase in bandwidth utilization rate will also cause problems such as video stuttering and slow page loading.

[0060] Therefore, fluctuation characteristic analysis can be performed according to the CPU utilization rate and bandwidth utilization rate. Further, in some embodiments of the present invention, according to the fluctuation discreteness and range of the CPU utilization rate and bandwidth utilization rate, determine the CPU peak fluctuation characteristic index and bandwidth peak fluctuation characteristic index of each server node, including: calculating the product of the variance and range of the CPU utilization rate at all sampling times of each server node, and performing normalization processing as the CPU peak fluctuation characteristic index; calculating the product of the variance and range of the bandwidth utilization rate at all sampling times of each server node, and performing normalization processing as the bandwidth peak fluctuation characteristic index.

[0061] The variance characterizes the discreteness of fluctuations. The larger the variance value, the greater the fluctuation of the corresponding CPU utilization. The larger the range value, the larger the overall distribution of CPU utilization values. Therefore, the more likely it is to have a peak effect. In this embodiment of the present invention, the product of the variance and the range of CPU utilization is directly calculated and normalized as a CPU peak fluctuation characteristic indicator. Similarly, the product of the variance and the range of bandwidth utilization is calculated and normalized as a bandwidth peak fluctuation characteristic indicator.

[0062] When the peak fluctuation performance of the CPU utilization and the peak fluctuation performance of the bandwidth utilization of the server node are greater, it means that there is obvious consumption of the CPU utilization and bandwidth utilization of the node, and the node may have a refresh peak period.

[0063] Therefore, in an embodiment of the present invention, the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index are combined to screen high-frequency refresh nodes from server nodes, including: calculating the product value of the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index, and normalizing it as a high-frequency refresh characteristic value; and taking a server node whose high-frequency refresh characteristic value is greater than a preset refresh characteristic threshold as a high-frequency refresh node.

[0064] The preset refresh feature threshold is a threshold value of the high-frequency refresh feature value. The specific value of the preset refresh feature threshold may be, for example, 0.5, that is, the server node with a high-frequency refresh feature value greater than 0.5 is regarded as a high-frequency refresh node.

[0065] S103: According to the frequency distribution changes and load data changes of user refresh requests at each sampling moment of the high-frequency refresh node, the high-frequency refresh moment is filtered out from the sampling moment of each high-frequency refresh node; according to the distribution of the high-frequency refresh moment in time series, the refresh segment with dense refresh distribution is determined, and according to the load changes of all high-frequency refresh moments in the refresh segment, the unconscious refresh performance evaluation value of the high-frequency refresh node is determined.

[0066] During peak periods such as live classes, online education platforms may encounter performance bottlenecks due to a large number of users accessing the platform at the same time. For example, video streaming freezes, page loading is slow, and request timeouts frequently occur, seriously affecting the user experience. When users encounter problems such as page loading failures or video playback interruptions, they frequently refresh the page and try to reload the content, resulting in a sharp increase in node requests.

[0067] When there are no problems such as video stream jams, slow page loading, or request timeouts on the user's page, the unconscious and frequent page refreshing behavior of users can also lead to a sharp increase in the request volume of nodes, increasing the system load. Without discrimination, these unconscious refresh requests and the requests of normal users will be confused, causing the load balancing algorithm to allocate too many requests to some nodes, thus creating new performance bottlenecks.

[0068] In addition, the resource consumption brought by these repeated requests may slow down the response speed of other normal requests, further affecting the overall performance of the platform. Therefore, it is necessary to distinguish between the jammed refresh moments and the unconscious refresh moments in the high-frequency refresh nodes, that is, it is necessary to conduct a specific analysis of the unconscious refresh performance.

[0069] Furthermore, in some embodiments of the present invention, according to the change in the frequency distribution of user refresh requests and the change in load data at each sampling moment of the high-frequency refresh nodes, high-frequency refresh moments are screened from the sampling moments of each high-frequency refresh node, including: determining the demand factor of the refresh request at each sampling moment according to the frequency distribution and discreteness characteristics of user refresh requests at all sampling moments of the same high-frequency refresh node; calculating the demand factor of the load data at each sampling moment of the same high-frequency refresh node by analogy with the demand factor of the refresh request; determining the refresh request volume-load dependence index of the refresh request and the load at each sampling moment according to the correlation and numerical difference between the demand factor of the refresh request and the demand factor of the load data at each sampling moment; taking the sampling moments with the refresh request volume-load dependence index greater than the preset dependence threshold as the high-frequency refresh moments.

[0070] Among them, the demand factor of the refresh request represents the degree of need for a refresh request at the corresponding sampling moment, that is, the larger the value of the demand factor of the refresh request, the more a refresh request is needed at the corresponding sampling moment. Specifically, it is manifested as: while the overall refresh request distribution is relatively stable, the refresh request volume is relatively large. Based on this, the calculation of the demand factor of the refresh request is specifically carried out.

[0071] Furthermore, in some embodiments of the present invention, according to the frequency distribution and discreteness characteristics of user refresh requests at all sampling moments of the same high-frequency refresh node, determining the demand factor of the refresh request at each sampling moment includes: taking any sampling moment as the target moment, calculating the mean value of the frequencies of user refresh requests at all sampling moments of the same high-frequency refresh node, normalizing the difference between the frequency of the user refresh request at the target moment and the mean value as the refresh frequency difference; calculating the negative value of the variance of the frequencies of user refresh requests at all sampling moments of the same high-frequency refresh node and performing maximum-minimum normalization to obtain the refresh frequency stability index; taking the ratio of the refresh frequency difference and the refresh frequency stability index as the demand factor of the refresh request at the target moment.

[0072] That is, by the difference between the frequency and the mean, the difference in the refresh frequency is obtained. The larger the value of the difference in the refresh frequency, the greater the difference between the frequency and the whole, and the higher the value, the more abnormal it is. The greater the demand factor of the refresh request, and the larger the value of the variance, the more unstable it is, and the lower the reliability of the obtained demand factor value. Calculate the stability index of the refresh frequency to represent the stability characteristics. Therefore, the ratio of the difference in the refresh frequency to the stability index of the refresh frequency is used as the demand factor of the refresh request at the target time.

[0073] Among them, the calculation method of the demand factor of the refresh request is similar to that of the demand factor of the load data. The specific calculation method of the demand factor of the load data is to calculate the mean value of the load data at all sampling times of the same high-frequency refresh node, and normalize the difference between the load data at the target time and the mean value as the load data difference; calculate the negative value of the variance of the load data at all sampling times of the same high-frequency refresh node, and perform maximum-minimum normalization to obtain the load stability index; use the ratio of the load data difference to the load stability index as the demand factor of the load data at the target time.

[0074] In the embodiments of the present invention, both the demand factor of the refresh request and the demand factor of the load data represent the characteristics of the refresh request and the load data at the corresponding sampling times. The correlation between the refresh request and the load change can be determined through correlation analysis, so as to determine the dependence of the refresh request on the load change.

[0075] Furthermore, in some embodiments of the present invention, according to the correlation and numerical difference between the demand factor of the refresh request and the demand factor of the load data at each sampling time, the refresh request volume-load dependence index between the refresh request and the load at each sampling time is determined, including: calculating the correlation index between the demand factor of the refresh request and the demand factor of the load data at each time based on the Pearson correlation coefficient; calculating the absolute value of the difference between the demand factor of the refresh request and the demand factor of the load data at any sampling time, and normalizing the negative value of the absolute value of the difference as the demand similarity index at the corresponding sampling time; calculating the product of the demand similarity index and the correlation index, and normalizing it to obtain the refresh request volume-load dependence index at the corresponding sampling time.

[0076] Among them, the Pearson correlation coefficient is a well-known correlation calculation method in the art. The correlation index is calculated, and the larger its value, the more positively correlated the fluctuations are.

[0077] After that, the larger the demand similarity index at the sampling moment is, the closer the numerical values of the demand factor of the refresh request and the demand factor of the load data are, which further indicates the similarity in terms of numerical values. Therefore, the product of the demand similarity index and the correlation index is calculated, and after normalization, the refresh request volume-load dependence index corresponding to the sampling moment is obtained. The larger the value of the refresh request volume-load dependence index is, the higher the similarity in both the numerical value and the fluctuation of the refresh request and the load data, that is, the greater the dependence of the refresh request on the load.

[0078] When each client refreshes the page, it will send a request to the server node to which it belongs, occupying certain system resources, such as CPU, memory, network bandwidth, etc. In the scenario of high-frequency refresh requests, the processing time for each request is relatively short, but frequent requests will cause the system to continuously occupy resources, resulting in an increase in load. For the live broadcast of an online education platform, if the frequency of refresh requests is too high, the server must continuously process these requests, which will correspondingly lead to an overloaded load; when the number of refresh requests is small, that is, in the case of non-high-frequency refresh, the load will be relatively small.

[0079] In the embodiment of the present invention, the refresh request volume-load dependence index is positively correlated with the demand factor of the refresh request and negatively correlated with the demand factor of the load data. That is, the larger the value of the refresh request volume-load dependence index is, it can further indicate that there are more refresh requests under low load conditions, which can be expressed as a high-frequency refresh situation. Therefore, the high-frequency refresh moment can be analyzed according to the refresh request volume-load dependence index.

[0080] The sampling moment when the refresh request volume-load dependence index is greater than the preset dependence threshold is used as the high-frequency refresh moment. Among them, the preset dependence threshold is the threshold value of the refresh request volume-load dependence index. In the embodiment of the present invention, the preset dependence threshold can be set to 0.8, that is, the sampling moment when the refresh request volume-load dependence index is greater than 0.8 is used as the high-frequency refresh moment.

[0081] The high-frequency refresh moment indicates the moment when the refresh far exceeds the influence of the current load. Then, the high-frequency refresh moment can be specifically segmented to determine the refresh segment that is greatly affected by the high-frequency refresh.

[0082] Further, in some embodiments of the present invention, according to the distribution of high-frequency refresh moments in time sequence, determining a refresh segment with dense refresh distribution includes: taking the time interval between any high-frequency refresh moment and the previous high-frequency refresh moment as the high-frequency interval corresponding to the high-frequency refresh moment; taking the other high-frequency intervals on both sides of any high-frequency interval in time sequence as adjacent intervals, and calculating the average duration of the adjacent intervals as the influence interval value of the middle high-frequency interval; calculating the ratio of the duration of the middle high-frequency interval to the influence interval value as a special ratio; when the special ratio is greater than a preset ratio, taking the middle high-frequency interval as a special interval; taking the time interval between the two nearest special intervals as the refresh segment with dense refresh distribution.

[0083] See Figure 3 , Figure 3 is a schematic diagram of a refresh segment provided by an embodiment of the present invention; the time interval between two high-frequency refresh moments is a high-frequency interval, and when the high-frequency interval is relatively large (referring to a longer duration compared with the left and right sides), it is taken as a special interval, and the time interval between two special intervals is taken as a refresh segment.

[0084] When specifically calculating the special interval, in order to specifically target the feature of "relatively large high-frequency interval", this solution introduces a feature ratio, and takes the high-frequency interval with a special ratio greater than the preset ratio as the special interval. The preset ratio can be specifically, for example, 3, that is, when the high-frequency interval is greater than 3 times the average value of the adjacent high-frequency intervals, the corresponding high-frequency interval is recorded as the special interval.

[0085] It should be noted that in this solution, the refresh is divided into two cases. One is the stuttering effect caused by the change of the load, resulting in the need for refresh, and the other is the unconscious refresh caused by the user's personal operation habits.

[0086] The stuttering refresh is usually accompanied by an increase in network or system load. The increase in refresh requests directly reflects the bottleneck problem of platform performance. Therefore, the relationship between the refresh request volume and the load is closer, and the load gradually rises as the request volume increases. The request volume of the user's unconscious page refresh does not match the load of the platform, that is, the requests may accumulate in a short time, but there is no obvious abnormal change in the system load. The refresh request volume-load dependence under the corresponding high-frequency refresh moment is less than that of the stuttering refresh. And the dependence of the refresh request on the load is greater under the stuttering refresh, and the dependence of the refresh request on the load is smaller under the unconscious refresh.

[0087] Further, in some embodiments of the present invention, an unconscious refresh performance evaluation value of a high-frequency refresh node is determined according to the load changes at all high-frequency refresh times within a refresh segment, including: calculating the mean value of the load data at all high-frequency refresh times within the refresh segment, and performing normalization processing to obtain the load coefficient of the refresh segment; taking the refresh segment with a load coefficient less than a preset load threshold as an unconscious segment; and determining the unconscious refresh performance evaluation value according to the number of high-frequency refresh times corresponding to the unconscious segment of the high-frequency refresh node.

[0088] Since some unconscious characteristics have been screened out by the value of the refresh request volume-load dependence index, it is necessary to further analyze the screened high-frequency refresh times. Among them, the load coefficient represents the numerical characteristics of the load. When the load is large, the refresh is more likely to be caused by lag, and when the load is small, the refresh is more likely to be an unconscious manifestation. Therefore, the refresh segment with a load coefficient less than the preset load threshold is taken as an unconscious segment. Among them, the preset load threshold is the threshold value of the load coefficient. In the embodiments of the present invention, the preset load threshold is set to 0.5, that is, the refresh segment with a load coefficient less than 0.5 is taken as an unconscious segment.

[0089] Then, the performance evaluation can be carried out in combination with the number of high-frequency refresh times in the unconscious segment. Further, in some embodiments of the present invention, an unconscious refresh performance evaluation value is determined according to the number of high-frequency refresh times corresponding to the unconscious segment of the high-frequency refresh node, including: taking the ratio of the number of high-frequency refresh times corresponding to the unconscious segment of the high-frequency refresh node to the number of all high-frequency refresh times as the unconscious refresh performance evaluation value.

[0090] The larger the ratio of the number of high-frequency refresh times corresponding to the unconscious segment to the number of all high-frequency refresh times, that is, the larger the proportion of unconscious refresh compared to lag refresh, the larger the value of the unconscious refresh performance evaluation value.

[0091] S104: Determine the unconscious influence nodes according to the unconscious refresh performance evaluation value of each server node, and limit the maximum refresh frequency of the unconscious influence nodes.

[0092] In the embodiments of the present invention, the unconscious refresh performance evaluation value can effectively characterize the unconscious effect. Therefore, the server nodes affected by unconsciousness can be directly determined. Determining the unconscious influence nodes according to the unconscious refresh performance evaluation value of each server node includes: taking the server nodes with an unconscious refresh performance evaluation value greater than the preset performance threshold as the unconscious influence nodes.

[0093] Among them, the preset performance threshold is the threshold value for evaluating the unconscious refresh performance. The preset performance threshold in the embodiments of the present invention can specifically be 0.5, that is, the server nodes with an unconscious refresh performance evaluation value greater than 0.5 are regarded as unconscious influence nodes. That is, when the number of unconscious refresh moments in each node is greater than half of the number of high-frequency refresh moments, the node is recorded as an unconscious influence node. An unconscious influence node indicates that unconscious refresh influence is likely to occur under the corresponding server node, so its maximum refresh frequency can be restricted, that is, its refresh frequency is reduced, thereby liberating this part of the cache resources, reducing the system burden, improving the overall resource utilization efficiency, and completing cloud information management.

[0094] For example, in a normal server node, its refresh times are not limited. When it is identified as an unconscious influence node, the refresh of the unconscious influence node is restricted by adopting conditions such as refreshing once every 3 seconds or only being able to refresh 10 times within 1 minute, so as to release the cache resources.

[0095] The present invention, through the frequency of refresh requests, load data, CPU utilization rate, and bandwidth utilization rate, takes the CPU utilization rate and bandwidth utilization rate as a group. By specifically analyzing the changes in the CPU state and bandwidth state, high-frequency refresh nodes are screened out from the server nodes; then, for the high-frequency refresh nodes, through the changes in the frequency distribution of refresh requests and load data, the high-frequency refresh moments are determined, that is, the high-frequency refresh moments of the unconscious refresh and stuttering refresh sets are initially screened out. Then, according to the distribution in time series and load changes of the high-frequency refresh moments, the unconscious refresh performance evaluation value of the high-frequency refresh nodes is determined. Through the unconscious refresh performance evaluation value, the unconscious influence nodes with relatively frequent unconscious refreshes can be accurately characterized, and then the maximum refresh frequency of the unconscious influence nodes is restricted, which can liberate this part of the cache resources, reduce the system burden, and improve the overall resource utilization efficiency.

[0096] It should be noted that: the above sequence of the embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0097] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. A cloud information management method for an online education platform, characterized in that: The method comprises: Obtain the frequency and load data of user refresh requests at each collection time of each server node of the online education platform, as well as multi-dimensional real-time data collected at all collection times of all server nodes, wherein the real-time data includes: CPU utilization and bandwidth utilization; Determine the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index of each server node according to the fluctuation discreteness and range of the CPU utilization rate and the bandwidth utilization rate; and select high-frequency refresh nodes from the server nodes by combining the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index; According to the frequency distribution changes and load data changes of user refresh requests at each sampling moment of the high-frequency refresh node, the high-frequency refresh moment is screened from the sampling moment of each high-frequency refresh node; according to the distribution of the high-frequency refresh moment in the time series, the refresh segment with dense refresh distribution is determined, and according to the load changes of all high-frequency refresh moments in the refresh segment, the unconscious refresh performance evaluation value of the high-frequency refresh node is determined; Determine the unconscious impact node according to the unconscious refresh performance evaluation value of each server node, and limit the maximum refresh frequency of the unconscious impact node; The step of determining the unconscious refresh performance evaluation value of the high-frequency refresh node according to the load changes at all high-frequency refresh moments in the refresh segment includes: Calculate the mean of the load data at all high-frequency refresh moments in the refresh segment, and perform normalization processing to obtain the load coefficient of the refresh segment; The refresh segment whose load factor is less than a preset load threshold is regarded as an unconscious segment; An unconscious refresh performance evaluation value is determined according to the number of high-frequency refresh moments corresponding to the unconscious segments of the high-frequency refresh nodes.

2. A cloud information management method for an online education platform as claimed in claim 1, characterized in that: Determining the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index of each server node according to the fluctuation discreteness and range of the CPU utilization and the bandwidth utilization includes: Calculate the product of the variance and range of the CPU utilization of each server node at all sampling times, and normalize them as the CPU peak fluctuation characteristic indicator; The product of the variance and range of the bandwidth utilization of each server node at all sampling times is calculated and normalized as the bandwidth peak fluctuation characteristic indicator.

3. The cloud information management method of an online education platform according to claim 1, characterized in that: The step of combining the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index to select high-frequency refresh nodes from server nodes includes: Calculate the product value of the CPU peak fluctuation characteristic index and the bandwidth peak fluctuation characteristic index, and normalize them as the high-frequency refresh characteristic value; The server node whose high-frequency refresh characteristic value is greater than a preset refresh characteristic threshold is regarded as a high-frequency refresh node.

4. The cloud information management method of an online education platform according to claim 1, characterized in that: The method of filtering out the high-frequency refresh time from the sampling time of each high-frequency refresh node according to the frequency distribution change of the user refresh request and the load data change at each sampling time of the high-frequency refresh node includes: Determine the demand factor of the refresh request at each sampling moment according to the frequency distribution and discrete characteristics of the user refresh request at all sampling moments of the same high-frequency refresh node; The demand factor of the load data at each sampling moment of the same high-frequency refresh node is calculated by analogy with the demand factor of the refresh request; Determine a refresh request quantity-load dependency index of the refresh request and the load at each sampling moment according to the correlation and value difference between the demand factor of the refresh request and the demand factor of the load data at each sampling moment; The sampling moment when the refresh request quantity-load dependency index is greater than a preset dependency threshold is used as the high-frequency refresh moment.

5. A cloud information management method for an online education platform as claimed in claim 4, characterized in that: The step of determining the demand factor of the refresh request at each sampling moment according to the frequency distribution and discrete characteristics of the user refresh requests at all sampling moments of the same high-frequency refresh node includes: Taking any sampling time as the target time, calculating the average frequency of user refresh requests at all sampling times of the same high-frequency refresh node, and normalizing the difference between the frequency of user refresh requests at the target time and the average value as the refresh frequency difference; Calculate the inverse of the variance of the frequency of user refresh requests at all sampling moments of the same high-frequency refresh node, and perform maximum and minimum value normalization to obtain the refresh frequency stability index; The ratio of the refresh frequency difference to the refresh frequency stability index is used as a demand factor for a refresh request at a target time.

6. A cloud information management method for an online education platform as claimed in claim 5, characterized in that: Determining the refresh request quantity-load dependency index of the refresh request and the load at each sampling moment according to the correlation and numerical difference between the demand factor of the refresh request and the demand factor of the load data at each sampling moment includes: Calculate the correlation index between the demand factor of the refresh request and the demand factor of the load data at each moment based on the Pearson correlation coefficient; Calculate the absolute value of the difference between the demand factor of the refresh request and the demand factor of the load data at any sampling time, and normalize the inverse of the absolute value of the difference as the demand similarity index at the corresponding sampling time; The product of the demand similarity index and the correlation index is calculated and normalized to obtain the refresh request quantity-load dependency index at the corresponding sampling time.

7. The cloud information management method of an online education platform according to claim 1, characterized in that: The step of determining the refresh segments with dense refresh distribution according to the distribution of high-frequency refresh moments in the time sequence includes: The time interval between any high-frequency refresh moment and the previous high-frequency refresh moment is used as the high-frequency interval of the corresponding high-frequency refresh moment; The other high-frequency intervals on both sides of any high-frequency interval in the time series are regarded as adjacent intervals, and the average duration of the adjacent intervals is calculated as the impact interval value of the middle high-frequency interval; the ratio of the duration of the middle high-frequency interval to the impact interval value is calculated as the special ratio; When the special ratio is greater than a preset ratio, the middle high-frequency interval is used as a special interval; and the time interval between two special intervals closest to each other is used as a refresh segment with dense refresh distribution.

8. A cloud information management method for an online education platform as claimed in claim 7, characterized in that: The step of determining the unconscious refresh performance evaluation value according to the number of high-frequency refresh moments corresponding to the unconscious segments of the high-frequency refresh nodes includes: The ratio of the number of high-frequency refresh moments corresponding to the unconscious segments of the high-frequency refresh nodes to the number of all high-frequency refresh moments is used as the unconscious refresh performance evaluation value.

9. The cloud information management method of an online education platform according to claim 1, characterized in that: The step of determining the unconscious impact node according to the unconscious refresh performance evaluation value of each server node includes: The server node whose unconscious refresh performance evaluation value is greater than a preset performance threshold is regarded as an unconscious influencing node.

Images