A server operation and maintenance management method and system for self-recommended operation
By obtaining and analyzing the operation logs and monitoring data of the server operation and maintenance management system, calculating operation priorities and using online prediction models to recommend operation and maintenance operations, the cumbersome operation and maintenance operations in the existing system are solved, and the operation and maintenance efficiency are improved. User experience.
Patent Information
- Application Number
- CN202411194925.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-29
AI Technical Summary
The existing server operation and maintenance management system fails to make full use of user operation habit data and historical data, resulting in cumbersome operation and maintenance-related functions, and users need to perform cumbersome repeated operations to analyze the problems.
By obtaining operation logs and monitoring data, performing data analysis and cleaning, calculating operation priority indicators, and using online prediction models to recommend operations, simplifying user operation and maintenance management processes.
It improves the efficiency and user experience of operation and maintenance management, reduces unnecessary operation steps, and provides intelligent recommendations to help users quickly find the functional portal they need.
Smart Images

Figure CN118708444B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of server operation data processing and cloud computing, and in particular to a server operation and maintenance management method and system for self-recommended operations. Background Art
[0002] When using the server operation and maintenance management system, after the user discovers an abnormal situation reported by the system, he needs to take corresponding actions based on his existing knowledge and experience.
[0003] The existing server operation and maintenance management system does not fully utilize and analyze the user's operating habit data and historical data. After the operation and maintenance personnel log in to the system, they still have to perform a series of tedious operations, and these operations are often repetitive. For example, when an application status abnormality is found on a server, the user will query the operation log of the application to analyze the problem; if the UI interface design of the server operation and maintenance management system is complex, the user may have to click through multiple levels of menus before finding frequently used functions, which increases the complexity of operations for the user.
[0004] In summary, the existing technology has the problem of cumbersome operation of system operation and maintenance related functions due to the inability to fully utilize historical data and integrate user habit analysis. Summary of the invention
[0005] The present invention provides a server operation and maintenance management method and system for self-recommended operation, so as to at least partially solve the problems existing in the above-mentioned background technology. Specifically, the technical problem solved by the present invention is realized by adopting the following technical solutions:
[0006] A server operation and maintenance management method for self-recommended operations, comprising:
[0007] Obtaining operation logs: Based on the system operation logs, through the resource management service, determine the information of each log audit and form a system log information collection;
[0008] Obtain monitoring data: Based on application and service information, determine each system monitoring information through data collection and alarm services, and form a system monitoring information collection;
[0009] System data analysis: Based on the system log information set and system monitoring information set, through data cleaning processing, the corresponding system log set and system monitoring data at each moment are retained, and a system log and monitoring data set is formed;
[0010] Determine recommended operation information: Based on the system log and monitoring data set, sort the recommended operations and determine the recommended operation information;
[0011] User operation recommendation: Based on the recommended operation information and the ranking results of the recommended operations, the recommended operations are recommended to relevant users;
[0012] The specific method of determining the recommended operation information is:
[0013] Operation priority ranking calculation: Based on the system log and monitoring data set, the priority of subsequent executable operations is calculated; based on the priority of each executable operation, the operations are sorted, and part or all of the sorted executable operations are taken as recommended operations, thereby determining the recommended operation information;
[0014] or;
[0015] Online operation prediction: Based on the system log and monitoring data set, the recommended operation information is determined through the system online prediction model.
[0016] Further, the calculation of the priority of the subsequent executable operation includes:
[0017] First priority indicator calculation: Calculate the first priority indicator based on the system log and the system monitoring data in the monitoring data set;
[0018] Second priority indicator calculation: calculating the second priority indicator based on the system log and the system log set in the monitoring data set;
[0019] Operation priority index calculation: Based on the first priority index and the second priority index, calculate the corresponding operation priority index of each specific operation.
[0020] Further, the first priority indicator calculation includes obtaining real-time monitoring parameters, and the real-time monitoring parameters include:
[0021] Based on the monitoring data set in the system log and the monitoring data set, a real-time monitoring data set is obtained and formed at real-time time intervals; the real-time monitoring data set includes CPU load Fields, memory usage Field, Request Connection Number field.
[0022] Furthermore, the first priority indicator calculation further includes calculating a real-time monitoring indicator, and the calculating of the real-time monitoring indicator includes:
[0023] Based on the real-time monitoring data set, the real-time CPU load indicator function is calculated through the real-time CPU load indicator function. ;
[0024] Based on the real-time monitoring data set, the real-time memory usage indicator is calculated through the real-time memory usage indicator function. ;
[0025] Based on the real-time monitoring data set, the real-time request connection number indicator is calculated through the real-time request connection number indicator function ;
[0026] The real-time CPU load indicator function:
[0027] ;
[0028] Said It is a real-time CPU load indicator;
[0029] Said To monitor the maximum value of CPU load data in real time;
[0030] Said The minimum value of CPU load data in the real-time monitoring data set;
[0031] Said is the standard value of CPU load;
[0032] The real-time memory usage indicator function:
[0033] ;
[0034] Said It is a real-time memory usage indicator;
[0035] Said To monitor the maximum value of memory usage in the data set in real time;
[0036] Said The minimum value of memory usage data in the real-time monitoring data set;
[0037] Said is the standard value of memory usage;
[0038] The real-time request connection number indicator function:
[0039] ;
[0040] Said It is the real-time request connection number indicator;
[0041] Said To monitor the average number of connections in the data set in real time;
[0042] Said The maximum value of the connection rating is set for real-time monitoring data;
[0043] Said The minimum number of connections required for real-time monitoring of data sets.
[0044] Further, the first priority indicator calculation further includes forming a first priority indicator, and the forming the first priority indicator includes:
[0045] Based on real-time CPU load indicator function , Real-time memory usage indicators , Real-time request connection number indicator , calculate the first priority index through the first priority index function ;
[0046] The first priority indicator function is:
[0047] ;
[0048] Said is a constant parameter determined based on experience.
[0049] Further, the second priority indicator calculation includes obtaining system log parameters, and the obtaining system log parameters includes:
[0050] Based on the system log and the system log set in the monitoring data set, the user information of the logged-in user is obtained, and the user history data set is formed by data extraction based on the user information;
[0051] Based on the user historical data set, an application historical data set is formed by extracting data based on specific application information;
[0052] Based on the application history data set, determine whether the number of records in the application history data set exceeds the history record number threshold. If it exceeds the history record number threshold, keep the data set record and form a basic application history data set. Otherwise, supplement the data set record and form a basic application history data set.
[0053] The obtaining of system log parameters also includes:
[0054] The supplementary data set record includes a difference supplementary record, and the difference supplementary record includes:
[0055] Get the number of records in the application history data set and count it as the current number of records;
[0056] Determine the difference between the current number of records and the threshold number of historical records;
[0057] The difference in records is used as the supplementary quantity, and the basic application history data set is formed after the application history data set is supplemented with the corresponding historical records;
[0058] The supplementary data set record includes an excess supplementary record, and the excess supplementary record includes:
[0059] Get the number of records in the application history data set and count it as the current number of records;
[0060] Determine the difference between the current number of records and the threshold number of historical records;
[0061] The multiple of the record difference is used as the supplementary quantity, and the basic application history data set is formed after the application history data set is supplemented with the corresponding historical records;
[0062] Based on the basic application history data set, by classifying the data based on the operation behavior, corresponding application history data subsets of different operation behaviors are formed; the corresponding number of each operation behavior historical records in the application history data subset of each operation behavior is counted respectively; the application history data subset of each operation behavior includes the operation behavior Fields, operation results Fields, request results in request data field, and count the number of data records in each subset , the total number of records in the historical log data is recorded as N;
[0063] Based on the system log and the monitoring data set in the monitoring data set, a historical monitoring data set is obtained at a certain time interval; the historical monitoring data set includes a historical CPU load field , Historical memory usage Field, historical request connection number field.
[0064] Furthermore, the second priority indicator calculation further includes calculating a historical monitoring indicator, and the calculating historical monitoring indicator includes:
[0065] Based on the historical monitoring data set, calculate the historical CPU load indicator function through the historical CPU load indicator function ;
[0066] Based on the historical monitoring data set, the historical memory usage indicator is calculated through the historical memory usage indicator function. ;
[0067] Based on the historical monitoring data set, the historical request connection number indicator is calculated through the historical request connection number indicator function ;
[0068] The historical CPU load indicator function:
[0069] ;
[0070] Said It is the historical CPU load indicator;
[0071] Said It is the maximum value of CPU load data in the historical monitoring data set;
[0072] Said It is the minimum value of CPU load data in the historical monitoring data set;
[0073] Said is the standard value of CPU load;
[0074] The historical memory usage indicator function:
[0075] ;
[0076] Said It is the historical memory usage indicator;
[0077] Said It is the maximum value of the memory usage data in the historical monitoring data set;
[0078] Said It is the minimum value of the memory usage data in the historical monitoring data set;
[0079] Said is the standard value of memory usage;
[0080] The historical request connection number indicators:
[0081] ;
[0082] Said This is the historical request connection number indicator;
[0083] Said is the average number of connections in the historical monitoring data set;
[0084] Said The maximum value of the connection rating requested for the historical monitoring data set;
[0085] Said Specifies the minimum number of connections requested for historical monitoring data.
[0086] Further, the second priority indicator calculation further includes forming a second priority indicator, wherein the forming of the second priority indicator includes: based on the application history data subset of each operation behavior, respectively calculating the second priority indicator of each operation behavior by using a second priority indicator function;
[0087] The second priority indicator function is:
[0088] ;
[0089] Said Any operational behavior;
[0090] Said The operation result corresponding to any operation behavior, which includes success, no response, and failure;
[0091] Said The request result corresponding to any operation behavior, which includes success, no response, and failure;
[0092] Said The number of data records for each action in the application history data subset;
[0093] The N is the total number of records in the historical log data.
[0094] Furthermore, the operation priority index calculation includes:
[0095] Based on the first priority indicator and the second priority indicator for each action , respectively, calculate the operation priority index of each operation behavior through the operation priority index function;
[0096] The operation priority index function is:
[0097] ;
[0098] Said It is an operation priority indicator.
[0099] Based on the operation priority index of each operation behavior, the operation priority index is sorted from high to low to determine the recommended operation information.
[0100] Further, the operation online prediction includes:
[0101] Based on the real-time system log and monitoring data training set, the system online prediction model generated by iterative training of the BP neural network algorithm determines the recommended operation information, where:
[0102] The input parameters of the system online prediction model are: real-time CPU load field , Real-time memory usage Field, real-time request connection number Fields, Operation Behavior Fields, operation results Fields, request results Field, standard value of CPU load Field, standard value of memory usage Field, average number of connections Fields;
[0103] The output of the system's online prediction model is: Operation behavior and The corresponding predicted probability.
[0104] At the same time, the present invention also provides a server operation and maintenance management system for self-recommendation operation, which is characterized by comprising:
[0105] The operation log acquisition module is used to determine the information of each log audit based on the system operation log through the resource management service and form a system log information collection;
[0106] The monitoring data acquisition module is used to determine each system monitoring information and form a system monitoring information set based on application and service information through data collection and alarm services;
[0107] System data analysis module: used to retain the corresponding system log set and system monitoring data at each moment through data cleaning processing based on the system log information set and the system monitoring information set, and form a system log and monitoring data set;
[0108] Determine recommended operation information module: sort the recommended operations based on the system log and monitoring data set, and determine the recommended operation information;
[0109] User operation recommendation module: based on the recommended operation information and the ranking results of the recommended operations, the recommended operations are recommended to relevant users;
[0110] The module for determining recommended operation information further includes:
[0111] Operation priority sorting calculation submodule: Calculate the priority of subsequent executable operations based on system logs and monitoring data sets; sort the operations based on the priority of each executable operation, and take part or all of the sorted executable operations as recommended operations, thereby determining recommended operation information;
[0112] or;
[0113] Operation online prediction submodule: Based on the system log and monitoring data set, the recommended operation information is determined through the system online prediction model.
[0114] Beneficial technical effects:
[0115] This solution sets up methods including obtaining operation logs, obtaining monitoring data, system data analysis, data offline training, data online prediction, user operation recommendation, etc., and recommends the operation entry required by the user on the homepage displayed after logging into the server operation and maintenance management system, and can directly click to jump to the corresponding function page. This solution can be displayed to the user according to the degree of recommendation (for example, sorting by color depth, not limited to form); this solution does not simply count the frequency of function use. It analyzes the monitoring data and log audit data of the operation and maintenance system based on the real-time operation status of the equipment, historical data, etc., through real-time priority calculation or machine learning algorithm. Therefore, the accuracy and efficiency of its intelligent recommendation are improved, which is more in line with the user's wishes, helping users save a lot of time and unnecessary operations, and improving the efficiency of operation and maintenance management. Usually, the monitoring data and log audit data of the computer operation and maintenance management system are basically stored in the database for a long time, except for providing necessary information for post-analysis and investigation and evidence collection, and are not fully utilized. Therefore, the operation and maintenance management solution provided by this solution can utilize this part of user historical operation data to improve the user's operation and maintenance experience. This solution can provide users with valuable operation suggestion entrances, thereby simplifying users' thinking and operation steps. BRIEF DESCRIPTION OF THE DRAWINGS
[0116] Figure 1 It is the main flow chart of the first embodiment of the operation and maintenance management method of the present invention;
[0117] Figure 2 It is a specific flow chart of the first embodiment of the operation and maintenance management method of the present invention;
[0118] Figure 3 It is a system flow chart of the first embodiment of the operation and maintenance management method of the present invention;
[0119] Figure 4 This is the main flow chart of the second embodiment of the operation and maintenance management method of the present invention;
[0120] Figure 5 It is a specific flow chart of the second embodiment of the operation and maintenance management method of the present invention;
[0121] Figure 6 It is a system flow chart of the second embodiment of the operation and maintenance management method of the present invention;
[0122] Figure 7 It is a BP neural network architecture diagram of the second embodiment of the operation and maintenance management method of the present invention;
[0123] Figure 8 BP neural network architecture diagram of the second embodiment of the operation and maintenance management method of the present invention;
[0124] Fig. 9It is a BP neural network model diagram of the second embodiment of the operation and maintenance management method of the present invention. DETAILED DESCRIPTION
[0125] The present invention is further described below in conjunction with the accompanying drawings:
[0126] In the figure:
[0127] S101-obtain operation log;
[0128] S102-obtain monitoring data;
[0129] S103-System data analysis;
[0130] S104-determine recommended operation information;
[0131] S105-User operation recommendation;
[0132] S1001-Based on the system operation log, through the resource management service, determine the information of each log audit and form a system log information collection;
[0133] S1002-Based on the application and service information, through data collection and alarm services, determine each system monitoring information and form a system monitoring information set;
[0134] S1003-Based on the system log information set and the system monitoring information set, through data cleaning processing, retain the corresponding system log set and system monitoring data at each moment, and form a system log and monitoring data set;
[0135] S1004- Sort the recommended operations based on the system log and monitoring data set to determine the recommended operation information;
[0136] S10041-Operation priority ranking calculation: Based on the system log and monitoring data set, the priority of subsequent executable operations is calculated; based on the priority of each executable operation, the operations are sorted, and part or all of the sorted executable operations are taken as recommended operations, thereby determining the recommended operation information;
[0137] S10042- Operation online prediction: Based on the system log and monitoring data set, the recommended operation information is determined through the system online prediction model;
[0138] S1005-Based on the recommended operation information, recommend the recommended operation to relevant users according to the ranking results of the recommended operations; Example
[0139] like Figure 1 , 2 , 3, a server operation and maintenance management method for self-recommended operations, comprising:
[0140] Obtaining operation logs S101: Based on the system operation logs, through the resource management service, determine the information of each log audit and form a system log information set S1001;
[0141] Acquiring monitoring data S102: Based on the application and service information, through data collection and alarm services, determining each piece of system monitoring information, and forming a system monitoring information set S1002;
[0142] System data analysis S103: Based on the system log information set and the system monitoring information set, through data cleaning processing, the corresponding system log set and system monitoring data at each moment are retained, and a system log and monitoring data set S1003 is formed;
[0143] Determine recommended operation information S104: Based on the system log and the monitoring data set, sort the recommended operations and determine recommended operation information S1004;
[0144] User operation recommendation S105: Based on the recommended operation information and according to the ranking result of the recommended operations, the recommended operations are recommended to relevant users S1005;
[0145] The specific method of determining the recommended operation information S104 is:
[0146] Operation priority ranking calculation: Based on the system log and monitoring data set, the priority of subsequent executable operations is calculated; based on the priority of each executable operation, the operations are ranked, and part or all of the ranked executable operations are taken as recommended operations, thereby determining recommended operation information S10041;
[0147] or;
[0148] Operation online prediction: Based on the system log and monitoring data set, the recommended operation information S10042 is determined through the system online prediction model.
[0149] This solution includes obtaining operation logs, obtaining monitoring data, system data analysis, determining recommended operation information, and user operation recommendations, wherein the specific method of determining recommended operation information includes: operation priority ranking calculation and operation online prediction. The above solution is divided into two parallel technical solutions: first, based on the various types of server data that have been obtained, the priority indicators of each operation and maintenance operation are directly calculated for recommendation, wherein the solution is to determine the recommended service through direct priority calculation and priority ranking; second, the operation and maintenance operation is recommended through the online prediction method of the neural network after offline training, wherein the recommended operation is determined through the online prediction method of the BP network model. In the homepage displayed after logging in to the server operation and maintenance management system, the operation entry required by the user will be recommended and displayed, and the user can directly click to jump to the corresponding function page. This solution can be displayed to the user according to the degree of recommendation (for example, sorting by color depth, not limited to form); this solution does not simply count the frequency of function use. It uses real-time priority calculation or machine learning algorithms (such as naive Bayes, artificial neural network, etc.) based on the reasonable analysis of historical data to analyze the monitoring data and log audit data of the operation and maintenance system. Therefore, the accuracy and efficiency of its intelligent recommendations are improved, which is more in line with the users' wishes, helping users save a lot of time and unnecessary operations, and improving operation and maintenance management efficiency.
[0150] Usually, the monitoring data and log audit data of the computer operation and maintenance management system are basically stored in the database for a long time, except for providing necessary information for post-analysis and investigation and evidence collection, and are not fully utilized. Therefore, the operation and maintenance management system provided by this solution can use this part of user historical data to improve the user's operation and maintenance operation experience. Based on the equipment operation status and historical data analysis technology, this solution provides users with valuable operation suggestion entrances, thereby simplifying users' thinking and operation steps.
[0151] The priorities of the subsequent executable operations of the calculation include:
[0152] First priority indicator calculation: Calculate the first priority indicator based on the system log and the system monitoring data in the monitoring data set;
[0153] Second priority indicator calculation: calculating the second priority indicator based on the system log and the system log set in the monitoring data set;
[0154] Operation priority index calculation: Based on the first priority index and the second priority index, calculate the corresponding operation priority index of each specific operation.
[0155] The main way to recommend priority indicators is to calculate the first priority indicator through the real-time monitoring data (for example, within 10 minutes); then calculate the second priority indicator by combining the data set formed by the historical log data and monitoring data (for example, the updated data set formed by the log data and monitoring data mentioned above). Based on the first priority indicator and the second priority indicator, the final priority index is calculated for each specific operation, and then the subsequent possible operations are sorted from high to low based on the priority index, and finally recommended to the user.
[0156] The first priority indicator calculation includes obtaining real-time monitoring parameters, and the real-time monitoring parameters include:
[0157] Based on the monitoring data set in the system log and the monitoring data set, a real-time monitoring data set is obtained and formed at real-time time intervals; the real-time monitoring data set includes CPU load Fields, memory usage Field, Request Connection Number field.
[0158] The log data and monitoring data used in the priority calculation algorithm are also obtained through the data set collection service. In the priority calculation, the data set formed by the obtained log data and monitoring data can also be obtained on a scheduled basis. For example, it can be obtained and updated on a scheduled basis every day, or at intervals of a certain number of days. Of course, if the device hardware conditions permit, the above-mentioned data set used in the priority calculation can also be obtained in real time, that is, when the operation recommendation service is started, the data set is obtained and formed in real time for use in the priority calculation.
[0159] In the implementation method using priority calculation, the cleaned data is added to the historical data to form a basic data set for priority calculation, thereby realizing the update of the data set. In actual operation, we can use the data in the most recent period of time in the data set as the data set for priority calculation. Since the data set is based on the time data in the log data and monitoring data as related information, we can use the time information of each data for filtering, such as filtering the data within the last three days as the data set for priority calculation, or within five days, etc.
[0160] After cleaning, each log data and monitoring data will form a piece of data according to the time point, and further constitute a data set. In the priority indicator calculation method of this embodiment, in the log data, we obtain the user information, specific application information, operation template, operation time, operation behavior, operation result, and request data; in the monitoring data, we obtain the CPU load, memory usage (or memory occupancy), and number of request connections.
[0161] User information is mainly used to identify specific users, so as to extract the corresponding cleaned log data and monitoring data; specific application information is used to determine the application the user is currently in, so as to filter out the corresponding application data set from the data set formed by the user's historical log data and monitoring data.
[0162] Real-time monitoring data is obtained. The real-time monitoring data can be, for example, data within 10 minutes or within 1 hour. The specific time range can be adjusted according to actual needs. The real-time monitoring data obtained includes CPU load, memory usage (or memory occupancy), and the number of requested connections. The above data is divided into time strips to form a real-time monitoring data set. The real-time monitoring data set can well reflect the current operating status, load status, etc. of the equipment. Due to the specific operations of subsequent maintenance personnel, it will have an impact on the equipment load, etc., and it also requires certain hardware resources as support. Therefore, the above data can provide a reference for various subsequent operation suggestions. In this embodiment, the first priority indicator will also be calculated based on the above real-time monitoring data set.
[0163] The first priority indicator calculation further includes calculating a real-time monitoring indicator, and the calculating of the real-time monitoring indicator includes:
[0164] Based on the real-time monitoring data set, the real-time CPU load indicator function is calculated through the real-time CPU load indicator function. ;
[0165] Based on the real-time monitoring data set, the real-time memory usage indicator is calculated through the real-time memory usage indicator function. ;
[0166] Based on the real-time monitoring data set, the real-time request connection number indicator is calculated through the real-time request connection number indicator function ;
[0167] The real-time CPU load indicator function:
[0168] ;
[0169] Said It is a real-time CPU load indicator;
[0170] Said To monitor the maximum value of CPU load data in real time;
[0171] Said The minimum value of CPU load data in the real-time monitoring data set;
[0172] Said is the standard value of CPU load;
[0173] The real-time memory usage indicator function:
[0174] ;
[0175] Said It is a real-time memory usage indicator;
[0176] Said To monitor the maximum value of memory usage in the data set in real time;
[0177] Said The minimum value of memory usage data in the real-time monitoring data set;
[0178] Said is the standard value of memory usage;
[0179] The real-time request connection number indicator function:
[0180] ;
[0181] Said It is the real-time request connection number indicator;
[0182] Said To monitor the average number of connections in the data set in real time;
[0183] Said The maximum value of the connection rating is set for real-time monitoring data;
[0184] Said The minimum number of connections required for real-time monitoring of data sets.
[0185] In server operation and maintenance, the three data indicators involved in real-time monitoring data often set corresponding standard operating status indicator values, or normal operating status indicator values, which we call standard values, as reference values for the optimal operating status. For example, the standard value of CPU load is The standard value of memory usage (or memory occupancy) is , the number of requested connections is generally set to a maximum recommended value, which we record as Based on the real-time monitoring data set, the indicators corresponding to CPU load, memory usage (or memory occupancy), and number of request connections are calculated as follows:
[0186] About CPU load indicator:
[0187] ;
[0188] in, Indicates the maximum value of the CPU load data in the real-time monitoring data set; Indicates the minimum value of the CPU load data in the real-time monitoring data set.
[0189] About calculating memory usage metrics:
[0190] ;
[0191] in, Indicates the maximum value of the memory usage data in the real-time monitoring data set; Indicates the minimum value of the memory usage data in the real-time monitoring data set.
[0192] About calculating the number of request connections:
[0193] In the calculation of the number of request connections, we first calculate the average number of connections in the real-time monitoring data set, that is, , then considering the number of server request connections and the stability requirements in server operation and maintenance, we calculate the number of request connections as follows:
[0194] ;
[0195] in, Indicates the average number of request connections in the real-time monitoring data set. Indicates the rated maximum value of request connections. In actual operation and maintenance, when the number of request connections exceeds the rated maximum value (that is, the maximum number of connections recommended for the server) for most of the time, the corresponding request connection number indicator may be calculated as a negative value, indicating that the server is in a state of request connection overload for a long time, which will affect subsequent operation and maintenance operations. For example, we may need to consider some subsequent request load transfer and scheduling-related operations. If its value is within the positive range, it means that the server has good resource redundancy in terms of request connections.
[0196] The first priority indicator calculation further includes forming a first priority indicator, wherein forming the first priority indicator includes:
[0197] Based on real-time CPU load indicator function , Real-time memory usage indicators , Real-time request connection number indicator , calculate the first priority index through the first priority index function ;
[0198] The first priority indicator function is:
[0199] ;
[0200] Said is a constant parameter determined based on experience.
[0201] Based on the above , and Calculate the first priority indicator The calculation of the first priority indicator is mainly used to evaluate the remaining consumable hardware resources for various operations based on the current server operation status, as a reference indicator for subsequent operations. In this context, the number of request connections and CPU load are often more important than memory usage. Therefore, the specific method of calculating the first priority indicator is as follows:
[0202] ;
[0203] The weight a is a constant parameter and can be determined based on empirical values. For example, it can be set to 0.4 or 0.6. Of course, a can also be set to 1.
[0204] The second priority indicator calculation includes obtaining system log parameters, and the obtaining system log parameters includes:
[0205] Based on the system log and the system log set in the monitoring data set, the user information of the logged-in user is obtained, and the user history data set is formed by data extraction based on the user information;
[0206] Based on the user historical data set, an application historical data set is formed by extracting data based on specific application information;
[0207] Based on the application history data set, determine whether the number of records in the application history data set exceeds the history record number threshold. If it exceeds the history record number threshold, keep the data set record and form a basic application history data set. Otherwise, supplement the data set record and form a basic application history data set.
[0208] The obtaining of system log parameters also includes:
[0209] The supplementary data set record includes a difference supplementary record, and the difference supplementary record includes:
[0210] Get the number of records in the application history data set and count it as the current number of records;
[0211] Determine the difference between the current number of records and the threshold number of historical records;
[0212] The difference in records is used as the supplementary quantity, and the basic application history data set is formed after the application history data set is supplemented with the corresponding historical records;
[0213] The supplementary data set record includes an excess supplementary record, and the excess supplementary record includes:
[0214] Get the number of records in the application history data set and count it as the current number of records;
[0215] Determine the difference between the current number of records and the threshold number of historical records;
[0216] The multiple of the record difference is used as the supplementary quantity, and the basic application history data set is formed after the application history data set is supplemented with the corresponding historical records;
[0217] Based on the basic application history data set, by classifying the data based on the operation behavior, corresponding application history data subsets of different operation behaviors are formed; the corresponding number of each operation behavior historical records in the application history data subset of each operation behavior is counted respectively; the application history data subset of each operation behavior includes the operation behavior Fields, operation results Fields, request results in request data field, and count the number of data records in each subset , the total number of records in the historical log data is recorded as N;
[0218] Based on the system log and the monitoring data set in the monitoring data set, a historical monitoring data set is obtained at a certain time interval; the historical monitoring data set includes a historical CPU load field , Historical memory usage Field, historical request connection number field.
[0219] After the user logs in, the user information is obtained. At the same time, the data set formed by the log data and monitoring data in the historical data is obtained, and the historical data of the corresponding user is extracted from it. The method of establishing the data set has been given above and will not be repeated here.
[0220] The specific application information of the user's current operation is obtained, and the historical data corresponding to the specific application is filtered from the above-mentioned historical data of the user to form application historical data.
[0221] Here, a more preferred way is that when we filter the application history data, we determine whether the number or amount of application history data corresponding to the application currently operated by the current user exceeds the preset data record number threshold. If it is greater than or equal to this threshold, the subsequent priority indicator calculation step is performed; if it is less than this threshold, the data corresponding to the current operation application other than the current user is supplemented from the historical data (i.e., the data set formed by the log data and the monitoring data). We call it supplementary historical data. The specific number of data records or amount of supplementary historical data can be determined according to actual needs. For example, the corresponding number of data records missing when the data record number threshold is reached can be supplemented. For example, if there is still a difference of 300 historical data from the data record number threshold, the number of supplementary historical data can be set to 300; more preferably, considering that the supplementary historical data is not the current user's own historical operation data, in order to reduce the data drift problem that may be caused in the calculation of the priority of the recommended operation, the number of supplementary historical data can be set to a certain multiple of the missing corresponding data record number, such as 1.2 times, 1.5 times, etc. of this difference. After the historical data is supplemented, complete application history data is formed.
[0222] Obtain application history data, and calculate the second priority index based on the application history data. In the calculation of the second priority index, we mainly focus on two parts, namely monitoring data, where we obtain CPU load, memory usage (or memory occupancy), number of request connections, and operation time, operation behavior, operation result, and request data in log data, where the operation behavior is the specific operation of the maintenance personnel on the current application in the current state, that is, the specific management operation of the application, and the operation behavior includes, for example, adding, deleting, updating, etc. The same operation behavior has the same corresponding data identifier in the log. Therefore, we can classify each piece of data in the application history data according to the operation behavior. In this embodiment, we record the operation behavior as, where m=1, 2, 3..., to distinguish each operation behavior, and form different application history data subsets based on the operation behavior, such as adding an operation subset, deleting an operation subset, etc. Then, count the number of data records in each application history data subset.
[0223] The second priority indicator calculation further includes calculating a historical monitoring indicator, and the calculating historical monitoring indicator includes:
[0224] Based on the historical monitoring data set, calculate the historical CPU load indicator function through the historical CPU load indicator function ;
[0225] Based on the historical monitoring data set, the historical memory usage indicator is calculated through the historical memory usage indicator function. ;
[0226] Based on the historical monitoring data set, the historical request connection number indicator is calculated through the historical request connection number indicator function ;
[0227] The historical CPU load indicator function:
[0228] ;
[0229] Said It is the historical CPU load indicator;
[0230] Said It is the maximum value of CPU load data in the historical monitoring data set;
[0231] Said It is the minimum value of CPU load data in the historical monitoring data set;
[0232] Said is the standard value of CPU load;
[0233] The historical memory usage indicator function:
[0234] ;
[0235] Said It is the historical memory usage indicator;
[0236] Said It is the maximum value of the memory usage data in the historical monitoring data set;
[0237] Said It is the minimum value of the memory usage data in the historical monitoring data set;
[0238] Said is the standard value of memory usage;
[0239] The historical request connection number indicators:
[0240] ;
[0241] Said This is the historical request connection number indicator;
[0242] Said is the average number of connections in the historical monitoring data set;
[0243] Said The maximum value of the connection rating requested for the historical monitoring data set;
[0244] Said Specifies the minimum number of connections requested for historical monitoring data.
[0245] For different subsets of application historical data, we calculate the second priority indicator. In the calculation of the second priority indicator, the data in the historical monitoring data is recorded as: historical CPU load , Historical memory usage , Historical request connection number ; In the historical log data, when calculating the second priority indicator, the data we focus on is recorded as: operation behavior , operation results, request results in the request data, and the number of data records in each subset , the total number of data records in the historical log data is recorded as N; in the historical log data, the operation time, that is, the timestamp of the log record, is mainly used to record the time point of the log, and is used to correspond, filter and store the log data with the monitoring data. Therefore, it is not used as a reference data for calculating the priority indicator here; in the request data, this embodiment mainly focuses on the result of the request, and the request source, request content, etc. are not used as the main reference data.
[0246] The operation results are assigned three values, namely 1, 0, and -1, which represent success, no response, and failure respectively. For the operation results, we focus on the success ratio, that is, for each application history data subset, we count the success rate of the operation results, which is recorded as , that is, the percentage of successful operations in the application history data subset.
[0247] The request result is set to three values, namely 1, 0, and -1, which represent success, no response, and failure respectively. Similarly, for the request result, we focus on the request success ratio, that is, for each application history data subset, we count the success rate of its request result, which is recorded as , that is, the percentage of successful requests in the application history data subset.
[0248] For historical CPU load , Historical memory usage , Historical request connection number , we refer to formulas (1)-(3) to calculate the corresponding historical CPU load indicators , Historical memory usage indicators And historical request connection number indicators , as follows:
[0249] ';
[0250] in, Indicates the maximum value of the historical CPU load data in the historical monitoring data set in a certain operation subset; Indicates the minimum value of the historical CPU load data in a subset of the historical monitoring data set. Indicates the standard value of CPU load.
[0251] Similarly, calculate the memory usage metric:
[0252] ;
[0253] in, Indicates the maximum value of memory usage data in the historical monitoring data set in a certain operation subset; Indicates the minimum value of memory usage data in a historical monitoring data set in a certain operation subset. Indicates the standard value of memory usage.
[0254] ;
[0255] in, Indicates the average number of request connections in a certain operation subset in the historical monitoring data set. Indicates the maximum value of the connection rating requested.
[0256] The second priority indicator calculation further includes forming a second priority indicator, wherein the forming the second priority indicator includes: based on the application history data subset of each operation behavior, respectively calculating the second priority indicator of each operation behavior by using a second priority indicator function;
[0257] The second priority indicator function is:
[0258] ;
[0259] Said Any operational behavior;
[0260] Said The operation result corresponding to any operation behavior, which includes success, no response, and failure;
[0261] Said The request result corresponding to any operation behavior, which includes success, no response, and failure;
[0262] Said The number of data records for each action in the application history data subset;
[0263] The N is the total number of records in the historical log data.
[0264] For operation behaviors in historical log data , Operation success rate , request result success rate in request data , and the number of data records in each subset , we combined historical CPU load indicators , Historical memory usage indicators And historical request connection number indicators , calculate the second priority index.
[0265] For a specific operation , its data subset is determined, then its second priority index The calculation is as follows:
[0266] ;
[0267] The second priority indicator is mainly used to calculate the recommended priority of each operation corresponding to the historical data, so as to give suggestions on specific operation recommendations based on the historical record data.
[0268] The operation priority index calculation includes:
[0269] Based on the first priority indicator and the second priority indicator for each action , respectively, calculate the operation priority index of each operation behavior through the operation priority index function;
[0270] The operation priority index function is:
[0271] ;
[0272] Said It is an operation priority indicator.
[0273] Based on the operation priority index of each operation behavior, the operation behaviors are sorted from high to low according to the operation priority index, thereby determining recommended operation information S104.
[0274] Based on the first priority index and the second priority index, calculate the Priority index. In the priority index calculation, we need to consider the influence of two aspects of parameters, namely the current state of the current application and the historical operation data. That is, the priority calculation in this embodiment is based on the first priority index and its parameters and the second priority index. Priority The specific calculation method is as follows:
[0275] ;
[0276] For different operations Priority , sort them from high to low according to the priority, and recommend the operations with higher priority to the operation and maintenance personnel.
[0277] At the same time, the present invention also provides a server operation and maintenance management system for self-recommendation operation, which is characterized by comprising:
[0278] The operation log acquisition module is used to determine the information of each log audit based on the system operation log through the resource management service and form a system log information set S1001;
[0279] The monitoring data acquisition module is used to determine each system monitoring information based on the application and service information through data collection and alarm services, and form a system monitoring information set S1002;
[0280] System data analysis module: used to retain the corresponding system log set and system monitoring data at each moment through data cleaning processing based on the system log information set and the system monitoring information set, and form a system log and monitoring data set S1003;
[0281] Determine recommended operation information module: based on the system log and monitoring data set, sort the recommended operations and determine the recommended operation information S1004;
[0282] User operation recommendation module: based on the recommended operation information and according to the ranking results of the recommended operations, recommend the recommended operations to relevant users S1005;
[0283] The module for determining recommended operation information further includes:
[0284] Operation priority ranking calculation submodule: based on the system log and monitoring data set, calculate the priority of subsequent executable operations; sort the operations based on the priority of each executable operation, and take part or all of the sorted executable operations as recommended operations, thereby determining recommended operation information S10041;
[0285] or;
[0286] Operation online prediction submodule: Based on the system log and monitoring data set, the recommended operation information is determined through the system online prediction model S10042.
[0287] This solution is set up to include an operation log acquisition module, a monitoring data acquisition module, a system data analysis module, a recommended operation information determination module, a user operation recommendation module, an operation priority ranking calculation submodule, an operation online prediction submodule, etc., wherein the specific method of determining the recommended operation information module includes two recommendation methods: operation priority ranking calculation and operation online prediction. The above scheme is divided into two parallel technical schemes: first, based on the various types of server data that have been acquired, recommendations are made by directly calculating the priority indicators of each operation and maintenance operation, wherein this scheme determines the recommended service through direct priority calculation and priority ranking; second, the operation and maintenance operation is recommended by online prediction through a neural network after offline training, wherein the recommended operation is recommended and determined through the online prediction method of the BP network model. The operation and maintenance management solution provided by this scheme can use historical operation data to improve the user's operation and maintenance operation experience. This scheme can provide users with valuable operation suggestion entrances based at least in part on user habit analysis, thereby simplifying the user's thinking and operation steps.
[0288] Embodiment 2:
[0289] A server operation and maintenance management method for self-recommended operations, comprising:
[0290] Under the determination of the recommended operation information scheme, an operation online prediction method is adopted;
[0291] Determine recommended operation information: Based on the system log and monitoring data set, sort the recommended operations and determine the recommended operation information;
[0292] Online prediction of operations: Based on system logs and monitoring data sets, the recommended operation information is determined through the system online prediction model.
[0293] like Figure 4 , 5 As shown in Figure 6, this solution provides a server operation and maintenance management system that quickly recommends operation entries. To improve the user's operation and maintenance experience, this solution provides users with valuable operation suggestion entries based on user habit analysis technology, thereby simplifying the user's thinking and operation steps.
[0294] Specifically, in this preferred embodiment, this solution can be implemented in the following specific ways:
[0295] Acquisition of monitoring data: Acquisition of monitoring data, which is stored in the monitoring subsystem of its computer operation and maintenance system. The monitoring subsystem usually consists of data collection, data storage and data query, and can integrate alarm services. The monitoring subsystem can be implemented by open source monitoring system software or existing technologies. The specific design of the monitoring subsystem is not the invention point of this solution;
[0296] System data analysis: Acquisition of operation log data: User operation data is stored in the log audit subsystem of its computer operation and maintenance system. When users of the operation and maintenance system operate their operation and maintenance management system, all management operations on resources (such as adding, deleting, and updating applications, etc.) will be recorded in the log audit subsystem. Each log audit information will include user information, function template, operation time, operation behavior, request data, operation results, etc. The information storage of the log audit subsystem can be achieved through MySQL, for example;
[0297] System data analysis: data collection, monitoring data comes from the operating system, application services, device resources, etc., and will change over time, such as CPU load, memory usage, and number of request connections. Data storage, monitoring data is saved in the local disk (or remote storage) through a time series database; data query, query the monitoring data through a data query language (such as PromQL);
[0298] The online prediction operation includes:
[0299] Based on the system log and monitoring data training set acquired in real time, the system online prediction model generated by iterative training of the BP neural network algorithm is used to determine the recommended operation information S104, wherein:
[0300] The input parameters of the system online prediction model are: real-time CPU load field , Real-time memory usage Field, real-time request connection number Fields, Operation Behavior Fields, operation results Fields, request results Field, standard value of CPU load Field, standard value of memory usage Field, average number of connections Fields;
[0301] The output of the system's online prediction model is: Operation behavior and The corresponding predicted probability.
[0302] This is achieved by using a neural network prediction method, that is, by building an offline model and using an online prediction method to calculate the prediction probability of each operation, sorting them by probability, and recommending them to the operation and maintenance personnel. The specific method is as follows:
[0303] Based on the real-time system log and monitoring data training set, the system online prediction model generated by iterative training of BP neural network algorithm determines the recommended operation information;
[0304] Before the online prediction model is generated, it also includes offline data training. Only through offline data training can the online prediction model be generated. The method is: based on the historical system log and monitoring data set and the historical operation behavior summary set, through the iterative training of the BP neural network algorithm, the offline training model is updated to generate the online prediction model;
[0305] The system log and monitoring data set includes the historical CPU load field , Historical memory usage Field, historical request connection number Fields, Operation Behavior Fields, operation results Fields, request results Field, standard value of CPU load Field, standard value of memory usage Field, average number of connections Historical monitoring datasets for fields;
[0306] The operation behavior summary set includes each operation behavior Fields and operation results corresponding to each operation behavior Fields and results of each operation The corresponding results Historical operation dataset of the field;
[0307] Finally, determine whether the offline training model has converged. If so, perform online prediction.
[0308] User operation recommendation: Operation suggestion interface: After the user logs in to the computer operation and maintenance system, the user information, application running status and other conditions are input to the back-end operation suggestion interface through HTTP request. The interface inputs the parameters into the priority calculation module, or into the online prediction service module of the machine learning algorithm module, and then recommends the functions that need to be operated to the user based on the user's historical operation habits and current operation conditions. The functions that need to be operated will be sorted according to the prediction probability or priority index, and the functions with the greatest possibility or the highest priority will be ranked first. The interface returns the recommendation results to the front end, and the server operation and maintenance management system displays the recommendation results to the user.
[0309] Data collection service: Based on local or remote application and service information, and according to the time series of applications and services, the system monitoring information is determined through data collection, data storage, and data query.
[0310] Since the data acquisition service includes: based on local or remote application and service information, according to the time series of applications and services, the system monitoring information is determined in sequence through data collection, data storage, and data query. Since this part of the data is stored in the monitoring subsystem of its computer operation and maintenance system. The monitoring subsystem is usually composed of data collection, data storage, and data query, and can integrate alarm services. The monitoring subsystem can be implemented by, for example, monitoring system software Prometheus, (1) The machine learning algorithm module regularly obtains the data of the training set, and the regular time can be every day, every three days, or every week. The shorter the cycle, the more timely the model update. (2) The data of the training set, namely the above-mentioned log data and monitoring data (including time dimension information, that is, all time series data), provides a data training set for the offline training part. After the machine learning algorithm module periodically obtains the log data and monitoring data in batches, it performs data cleaning, discards useless data, and retains the log data and monitoring data at the same time.
[0311] Data cleaning processing includes:
[0312] Filter out the system log information within the statistical time after the user logs in, and delete the redundant system log information beyond the statistical time;
[0313] Query the system monitoring information within the statistical time and delete the redundant system monitoring information beyond the statistical time;
[0314] According to the target results of the operation records and the conditional parameters of the monitoring data, a system log information set and a system monitoring information set are formed.
[0315] By using the above data cleaning work, useless data can be discarded, and the log data and monitoring data at the same time can be retained. Specifically, during data cleaning, the log audit data within a period of time after the user logs in is screened out, such as data within 10 minutes, and all data beyond this time is discarded; the monitoring data with the most recent timestamp corresponding to the query operation record, such as the server resource monitoring data with the most recent time corresponding to the query application details operation record (monitoring alarm level, reason, number of times, etc., such as severity level, high CPU, memory, disk IO usage, application restart, etc., 3 times), and all other monitoring data are discarded.
[0316] In an implementation method using an offline model for prediction, the cleaned training data is input into the algorithm module of the offline training part for training, and the training model (model file or model data stored in a database) is updated. The operation record is used as the target result, and the monitoring data is used as the conditional parameter. The combination of them is temporarily stored in the training set as a valid training data, and the training set is stored in a message queue or database.
[0317] In the implementation method using priority calculation, the cleaned data is added to the historical data to form a basic data set for priority calculation, thereby realizing the update of the data set. In actual operation, we can use the data in the most recent period of time in the data set as the data set for priority calculation. Since the data set is based on the time data in the log data and monitoring data as related information, we can use the time information of each data for filtering, such as filtering the data within the last three days as the data set for priority calculation, or within five days, etc.
[0318] In the specific operation recommendation, in another implementation, we can also use a neural network prediction method to implement it, that is, through offline model establishment and online prediction, calculate the prediction probability of each different operation, sort it by probability, and recommend it to the operation and maintenance personnel. The specific method is as follows:
[0319] The data offline training also includes:
[0320] Based on the cleaned system logs and monitoring data sets, the offline training model is updated through iterative training of the BP neural network algorithm;
[0321] If the expected output parameters and the actual output parameters meet the accuracy requirements, the iterative training ends; otherwise, the weights of the corresponding nodes in the hidden layer of the BP neural network continue to be updated.
[0322] Since the use of data offline training also includes: based on the cleaned system log and monitoring data set, through the iterative training of the BP neural network algorithm, the offline training model is updated; if the expected output parameters and the actual output parameters meet the accuracy requirements, the iterative training is terminated, otherwise, the weights of the corresponding nodes in the hidden layer of the BP neural network are continued to be updated. Since the cleaned training data is input into the algorithm module of the offline training part for training, the offline training part reads these training data from the message queue or database, and after the algorithm training, the training model is updated (the model file is generated or the model data is saved in the database). The algorithm module is in a state of constantly improving the training model within a certain period (such as half a year or a year, depending on the size of the algorithm training set data and the time range of the data). Therefore, as the number of times users use the computer operation and maintenance system increases, the algorithm training set data obtained will be richer, the algorithm model will be more perfect, and the algorithm prediction / recommendation results will be more and more accurate; during the training process, the basic model algorithm of the BP neural network algorithm determines whether the expected output and the actual output meet the accuracy requirements. If they meet, the training is terminated, and if not, the weights in its hidden layer are continued to be updated.
[0323] like Figure 7 , 8As shown in FIG. 9 , in a preferred embodiment, the basic structure of the BP neural network includes an input layer, a hidden layer and an output layer; the basic structure of the BP neural network, the number of input and output nodes, the number of hidden layers, and the initial weights are determined by input parameters and output parameters;
[0324] The input parameters of the BP neural network include operation time and environmental parameters, and the output parameters include operation behavior;
[0325] The environmental parameters include CPU resource usage and application status;
[0326] The operation behavior includes application details and alarm events.
[0327] Since the basic structure of the BP neural network includes an input layer, a hidden layer and an output layer; the basic structure of the BP neural network, the number of input and output nodes, the number of hidden layers, and the initial weights are determined by input parameters and output parameters; the input parameters of the BP neural network include operation time and environmental parameters, and its output parameters include operation behavior; the environmental parameters include CPU resource usage and application status; the operation behavior includes application details and alarm events. Since the BP neural network algorithm uses input parameters and output parameter pairs to determine its basic structure, the number of input and output nodes, the number of hidden layers, the initial weights, etc. Figure 4 As shown, according to the operation time and environment parameters (CPU resource usage, etc.) in the input parameters, the operation behavior of the corresponding output parameter is obtained, which may also include the probability of the corresponding operation behavior.
[0328] like Figure 4 , 5 As shown in , 6, the BP neural network algorithm includes:
[0329] Based on the current neuron node of the neuron model, the output value of the current neuron node is determined through the node activation function according to the output values of all upper-layer neuron nodes, the weights of the current neuron node and all upper-layer neuron nodes, and the threshold of the current neuron node;
[0330] If the error between the output value of the current neuron node and the expected value of the current neuron node does not meet the set accuracy requirement, the weight of the hidden layer is corrected through the error calculation formula to continue training for multiple iterations.
[0331] The BP neural network algorithm includes: based on the current neuron node of the neuron model, according to the output value of all upper-layer neuron nodes, the weight of the current neuron node and all upper-layer neuron nodes and the threshold of the current neuron node, the output value of the current neuron node is determined through the node activation function; if the error between the output value of the current neuron node and the expected value of the current neuron node does not meet the set accuracy requirement, the weight of the hidden layer is corrected through the error calculation formula to continue training, and multiple iterations are performed, and the weight between node i and node j is set to W ij , the threshold of node j is b j , the output value of each node is X j , and the output value of each node is realized based on the output value of all nodes in the upper layer, the weights of the current node and all nodes in the upper layer, the threshold of the current node, and the activation function.
[0332] The node activation function is:
[0333] ;
[0334] Said is the weight between the current neuron node and another neuron node; f(x) represents the activation function;
[0335] Said is the threshold of the current neuron node;
[0336] Said is the output value of the current neuron node;
[0337] Said , For all upper layer neuron nodes.
[0338] In the BP neural network of this embodiment, the error calculation formula is set as:
[0339] ;
[0340] Said is the training result of the output layer neuron nodes;
[0341] Said is the target value of the neuron node in the output layer;
[0342] The w is the weight between the input layer and the hidden layer neuron nodes or between the output layer and the hidden layer neuron nodes;
[0343] The b is the threshold between the input layer and the hidden layer or between the output layer and the hidden layer, and n represents the total number of neuron nodes in the output layer.
[0344] Since the error calculation formula is used, that is, the server calls the computing resources to calculate according to the above formula, the output value can be obtained. If the error between the output value and the expected value does not meet the set accuracy requirements, the weights of the hidden layer will be corrected to continue training for multiple iterations. This process is called back propagation.
[0345] After the BP neural network algorithm training described above is completed, the weights between the various node connections in the neural network have been updated, that is, the update of the algorithm model has been completed, and the model can be used for actual online prediction.
[0346] The online prediction of data also includes:
[0347] Based on the updated offline training model;
[0348] The input parameters of the system online prediction model are: real-time CPU load field, real-time memory usage field, real-time request connection number field, operation behavior field, operation result field, request result field, standard value field of CPU load, standard value field of memory usage, and average connection number field;
[0349] Predicted operations: Based on the user's historical operating habits and current operating conditions, the functions that need to be operated are recommended to the user. The functions that need to be operated will be sorted according to the predicted probability, and the functions with the greatest possibility will be ranked at the top to form the most likely reference;
[0350] The output of the system's online prediction model is: operation behavior and corresponding prediction probability.
[0351] Specifically, this embodiment uses a method based on an updated offline training model for prediction; parameter input: inputting user information and parameters related to the application running state into the updated offline training model; predicting operation: based on historical data and current operating conditions (such as parameters related to the current device running state, etc.), recommending functions that need to be operated to the user, the functions that need to be operated will be sorted according to the predicted probability, and the functions with the greatest possibility will be ranked at the forefront and become the most likely reference; reference output: outputting the user based on the most likely reference, the specific process of the online prediction part is as follows:
[0352] 1. Access the operation suggestion interface: When the user logs in to the computer operation and maintenance system, the front end automatically inputs user information, application running status and other conditions through HTTP request and sends them to the back-end operation suggestion interface; 2. The operation suggestion interface inputs the parameters into the online prediction service module of the machine learning algorithm module, and then recommends the functions that need to be operated to the user based on the user's historical operation habits and current operation conditions. The functions that need to be operated will be sorted according to the predicted probability, and the functions with the greatest possibility will be ranked first; 3. The interface results are returned to the front end. After parsing the results, the server operation and maintenance management system will display the recommended results to the user.
[0353] Utilizing the technical solution of the present invention, or those skilled in the art designing similar technical solutions inspired by the technical solution of the present invention, and achieving the above technical effects, all fall within the protection scope of the present invention. It should be pointed out that the specific implementations in the embodiments of the present invention can be at least partially combined with each other, and the combination and use of the above embodiments, or their different combinations, should also be regarded as falling within the protection scope of the present invention.
Claims
1. A server operation and maintenance management method for self-recommendation operation, characterized in that: include: Obtaining operation logs: Based on the system operation logs, through the resource management service, determine the information of each log audit and form a system log information collection; Obtain monitoring data: Based on application and service information, determine each system monitoring information through data collection and alarm services, and form a system monitoring information collection; System data analysis: Based on the system log information set and system monitoring information set, through data cleaning processing, the corresponding system log set and system monitoring data at each moment are retained, and a system log and monitoring data set is formed; Determine recommended operation information: Based on the system log and monitoring data set, sort the recommended operations and determine the recommended operation information; User operation recommendation: Based on the recommended operation information and the ranking results of the recommended operations, the recommended operations are recommended to relevant users; The specific method of determining the recommended operation information is: Operation priority ranking calculation: Based on the system log and monitoring data set, the priority of subsequent executable operations is calculated; based on the priority of each executable operation, the operations are sorted, and part or all of the sorted executable operations are taken as recommended operations, thereby determining the recommended operation information; or; Online operation prediction: Based on the system log and monitoring data set, the recommended operation information is determined through the system online prediction model; The priorities of the subsequent executable operations of the calculation include: First priority indicator calculation: Calculate the first priority indicator based on the system log and the system monitoring data in the monitoring data set; Second priority indicator calculation: calculating the second priority indicator based on the system log and the system log set in the monitoring data set; Operation priority index calculation: Based on the first priority index and the second priority index, calculate the corresponding operation priority index of each specific operation.
2. The server operation and maintenance management method according to claim 1, characterized in that: The first priority indicator calculation includes obtaining real-time monitoring parameters, and the real-time monitoring parameters include: Based on the monitoring data set in the system log and the monitoring data set, a real-time monitoring data set is obtained and formed at real-time time intervals; the real-time monitoring data set includes CPU load Fields, memory usage Field, Request Connection Number field.
3. The server operation and maintenance management method according to claim 2, characterized in that: The first priority indicator calculation further includes calculating a real-time monitoring indicator, and the calculating of the real-time monitoring indicator includes: Based on the real-time monitoring data set, the real-time CPU load indicator function is calculated through the real-time CPU load indicator function. ; Based on the real-time monitoring data set, the real-time memory usage indicator is calculated through the real-time memory usage indicator function. ; Based on the real-time monitoring data set, the real-time request connection number indicator is calculated through the real-time request connection number indicator function ; The real-time CPU load indicator function: ; Said It is a real-time CPU load indicator; Said To monitor the maximum value of CPU load data in real time; Said The minimum value of CPU load data in the real-time monitoring data set; Said is the standard value of CPU load; The real-time memory usage indicator function: ; Said It is a real-time memory usage indicator; Said To monitor the maximum value of memory usage in the data set in real time; Said The minimum value of memory usage data in the real-time monitoring data set; Said is the standard value of memory usage; The real-time request connection number indicator function: ; Said It is the real-time request connection number indicator; Said To monitor the average number of connections in a real-time data set; Said The maximum value of the connection rating is set for real-time monitoring data; Said The minimum number of connections required for real-time monitoring of data sets.
4. The server operation and maintenance management method according to claim 3, characterized in that: The first priority indicator calculation further includes forming a first priority indicator, wherein forming the first priority indicator includes: Based on real-time CPU load indicator function , Real-time memory usage indicators , Real-time request connection number indicator , calculate the first priority index through the first priority index function ; The first priority indicator function is: ; Said is a constant parameter determined based on experience.
5. The server operation and maintenance management method according to claim 1, characterized in that: The second priority indicator calculation includes obtaining system log parameters, and the obtaining system log parameters includes: Based on the system log and the system log set in the monitoring data set, the user information of the logged-in user is obtained, and a user history data set is formed by extracting data based on the user information; Based on the user historical data set, an application historical data set is formed by extracting data based on specific application information; Based on the application history data set, determine whether the number of records in the application history data set exceeds the history record number threshold. If it exceeds the history record number threshold, keep the data set record and form a basic application history data set. Otherwise, supplement the data set record and form a basic application history data set. The obtaining of system log parameters also includes: The supplementary data set record includes a difference supplementary record, and the difference supplementary record includes: Get the number of records in the application history data set and count it as the current number of records; Determine the difference between the current number of records and the threshold number of historical records; The difference in records is used as the supplementary quantity, and the basic application history data set is formed after the application history data set is supplemented with the corresponding historical records; The supplementary data set record includes an excess supplementary record, and the excess supplementary record includes: Get the number of records in the application history data set and count it as the current number of records; Determine the difference between the current number of records and the threshold number of historical records; The multiple of the record difference is used as the supplementary quantity, and the basic application history data set is formed after the application history data set is supplemented with the corresponding historical records; Based on the basic application history data set, by classifying the data based on the operation behavior, corresponding application history data subsets of different operation behaviors are formed; the corresponding number of each operation behavior historical records in the application history data subset of each operation behavior is counted respectively; the application history data subset of each operation behavior includes the operation behavior Fields, operation results Fields, request results in request data field, and count the number of data records in each subset , the total number of records in the historical log data is recorded as N; Based on the system log and the monitoring data set in the monitoring data set, a historical monitoring data set is obtained at a certain time interval; the historical monitoring data set includes a historical CPU load field , Historical memory usage Field, historical request connection number field.
6. The server operation and maintenance management method according to claim 5, characterized in that: The second priority indicator calculation further includes calculating a historical monitoring indicator, and the calculating historical monitoring indicator includes: Based on the historical monitoring data set, calculate the historical CPU load indicator function through the historical CPU load indicator function ; Based on the historical monitoring data set, the historical memory usage indicator is calculated through the historical memory usage indicator function. ; Based on the historical monitoring data set, the historical request connection number indicator is calculated through the historical request connection number indicator function ; The historical CPU load indicator function: ; Said It is the historical CPU load indicator; Said It is the maximum value of CPU load data in the historical monitoring data set; Said It is the minimum value of CPU load data in the historical monitoring data set; Said is the standard value of CPU load; The historical memory usage indicator function: ; Said It is the historical memory usage indicator; Said It is the maximum value of the memory usage data in the historical monitoring data set; Said It is the minimum value of the memory usage data in the historical monitoring data set; Said is the standard value of memory usage; The historical request connection number indicators: ; Said This is the historical request connection number indicator; Said is the average number of connections in the historical monitoring data set; Said The maximum value of the connection rating requested for the historical monitoring data set; Said Specifies the minimum number of connections requested for historical monitoring data.
7. The server operation and maintenance management method according to claim 6, characterized in that: The second priority indicator calculation further includes forming a second priority indicator, wherein the forming the second priority indicator includes: based on the application history data subset of each operation behavior, respectively calculating the second priority indicator of each operation behavior by using a second priority indicator function; The second priority indicator function is: ; Said Any operational behavior; Said The operation result corresponding to any operation behavior, which includes success, no response, and failure; Said The request result corresponding to any operation behavior, which includes success, no response, and failure; Said The number of data records for each action in the application history data subset; The N is the total number of records in the historical log data.
8. The server operation and maintenance management method according to claim 1, characterized in that: The operation priority index calculation includes: Based on the first priority indicator and the second priority indicator for each action , respectively, calculate the operation priority index of each operation behavior through the operation priority index function; The operation priority index function is: ; Said It is the operational priority indicator; Based on the operation priority index of each operation behavior, the operation priority index is sorted from high to low to determine the recommended operation information.
9. A server operation and maintenance management system with self-recommended operation, characterized in that: include: The operation log acquisition module is used to determine the information of each log audit based on the system operation log through the resource management service and form a system log information collection; The monitoring data acquisition module is used to determine each system monitoring information and form a system monitoring information set based on application and service information through data collection and alarm services; System data analysis module: used to retain the corresponding system log set and system monitoring data at each moment through data cleaning processing based on the system log information set and the system monitoring information set, and form a system log and monitoring data set; Determine recommended operation information module: sort the recommended operations based on the system log and monitoring data set, and determine the recommended operation information; User operation recommendation module: based on the recommended operation information and the ranking results of the recommended operations, the recommended operations are recommended to relevant users; The module for determining recommended operation information further includes: Operation priority sorting calculation submodule: Calculate the priority of subsequent executable operations based on system logs and monitoring data sets; sort the operations based on the priority of each executable operation, and take part or all of the sorted executable operations as recommended operations, thereby determining recommended operation information; or; Operation online prediction submodule: Based on the system log and monitoring data set, the recommended operation information is determined through the system online prediction model.
