Adaptive service alarm system based on multiple tenants

Abstract

The invention discloses a multi-tenant-based adaptive service alarm system, which relates to the field of service intelligent analysis and comprises a data collection module, an aggregation storage module, an alarm rule module, an alarm calculation module and an alarm notification module. Key business and system performance indexes are monitored in real time through the lightweight data collector, effective feature screening and anomaly detection are carried out on data through the decision tree model, the data quality is improved, efficient grouping summarization of the key business indexes and the system performance indexes is achieved by designing a proper database table structure and indexes, and the data quality is improved. By calculating floating values and weights of tenants under different time scales, an alarm threshold value can be adjusted according to real-time flow changes, false alarm or missing alarm possibly caused by a fixed threshold value is avoided, alarm rules are graded, corresponding priorities are set, multi-channel notification is supported, and a notification callback mechanism is established to record a sending result, so that the alarm efficiency is improved. And it is ensured that high-priority alarms can attract attention in time and optimize resource use.

Description

Technical Field

[0001] The present invention relates to the field of business intelligent analysis, and in particular to a multi-tenant-based adaptive business alarm system. Background Art

[0002] With the development of cloud computing, microservice architecture and big data technology, the IT infrastructure in online ride-hailing services has become increasingly complex and diverse. Many businesses have become highly dependent on the normal operation of information technology systems. Any system failure or anomaly may have a serious impact on the business, and more and more online ride-hailing service providers are beginning to seek real-time monitoring and alarm systems as a solution. This system can achieve real-time monitoring, anomaly detection and alarm notification of online ride-hailing services, helping online ride-hailing service providers to take action at the early stage of problems, thereby ensuring stable operation and continuous improvement of the business. Automated monitoring reduces the need for human intervention and related costs, improves efficiency, and enhances the ability to respond to emergencies.

[0003] The existing technology has the following shortcomings: the thresholds of existing monitoring alarms are usually preset and cannot adapt to the ever-changing business environment. Too static thresholds may lead to false alarms or missed alarms, affecting the response efficiency of the system. In microservice architectures and multi-cloud environments, each service may use different monitoring tools and technology stacks, which makes the integration and analysis of monitoring data complicated. The single alarm channel of existing technologies can lead to information overload problems. Due to excessive reliance on rules and thresholds and lack of intelligent analysis capabilities based on machine learning, it is difficult for the system to automatically adjust thresholds. When under high load, the alarm system may face performance bottlenecks, resulting in loss or delay of monitoring data, and failure to notify relevant personnel in a timely manner, thus missing the best time to solve the problem.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-tenant-based adaptive service alarm system to solve the problems in the above-mentioned background technology.

[0006] In order to achieve the above-mentioned object, the present invention provides the following technical solutions: an adaptive business alarm system based on multi-tenants, comprising a data collection module, an aggregation storage module, an alarm rule module, an alarm calculation module and an alarm notification module;

[0007] Data collection module: Set up a lightweight data collector to collect key business indicators and system performance indicators in real time, use the decision tree model to perform effective feature screening and anomaly detection, and perform cleaning and mean filling;

[0008] Aggregation storage module: Design a database table structure containing tenant ID and city index through MySQL, use GROUP BY aggregation function to group and summarize key business indicators and system performance indicators, implement ETL process and batch writing, and regularly store processing results;

[0009] Alarm rule module: Design alarm rule configuration database table through MySQL, provide API interface to submit and update alarm rules, and set alarm trigger mechanism for response time, throughput, error rate, number of user registrations, customer retention rate and average order value;

[0010] Alarm calculation module: Build a recurrent neural network model based on LSTM units, output the best estimated range of the alarm rules corresponding to the current tenant and city, and determine the tenant weight and floating value based on the traffic data at different time scales to determine the threshold of the corresponding alarm rule;

[0011] Alarm notification module: By dividing alarm rules into levels and setting corresponding priorities, it supports multi-channel notifications and establishes a notification callback mechanism to record the sending results.

[0012] Preferably, the car rental application log, car rental server performance, database, and network traffic are selected as data sources, the operation log of the car rental application is recorded through a centralized log management tool, a lightweight data collector is set to collect key business indicators and system performance indicators in real time, the RESTful API interface is used for status interaction with the car rental application log, the JDBC connection is used for interaction with the car rental server performance and database, and the network traffic is analyzed and interacted through a network monitoring tool.

[0013] Preferably, a tree structure is constructed, and the branch nodes inside the tree structure represent the invalid features of key business indicators and system performance indicators, and the leaf nodes inside the tree structure represent the invalid removal results. Multiple decision trees are constructed by randomly selecting invalid features, key business indicators and system performance indicators, and the invalid features of branch nodes are randomly selected according to the leaf nodes for optimal splitting. Independent classification prediction is performed on each decision tree, and the classification prediction that appears most frequently in all decision trees is selected as the invalid classification. The invalid key business indicators and system performance indicators are judged and removed, and duplicate key business indicators and system performance indicators are checked and deleted. It is checked whether the key business indicators and system performance indicators have null values ​​and missing values, and whether the key business indicators and system performance indicators are within the valid range. It is checked whether the key business indicators and system performance indicators have abnormalities and duplicate records through the constraints of uniqueness and consistency. For different existing situations, themes are determined and divided into different theme partitions, and the corresponding log data objects are deleted through each partition, and the key business indicators and system performance indicators are filled accordingly using the mean filling method.

[0014] Preferably, the key business indicators and system performance indicators are grouped according to the tenants and cities according to the data source, and the aggregation of key business indicators and system performance indicators is realized by using the GROUP BY aggregation function. MySQL is selected as the storage system, and the database table structure is designed to include the fields of key business indicators and system performance indicators. Indexes are created with tenant ID and city, and the aggregation of key business indicators and system performance indicators is stored regularly through batch writing and designing ETL process.

[0015] Preferably, define alarm rule parameters and alarm rules, select MySQL as the storage system, design the database table structure, provide an API interface to allow submission of alarm rule configurations, provide an update interface to allow modification of existing alarm rules, support querying alarm rules by tenant ID through the API interface, use the background service to set the alarm trigger mechanism, and use the message queue to periodically check whether the key business indicators and system performance indicators corresponding to each tenant meet the alarm rules.

[0016] Preferably, an LSTM unit is selected to construct a recurrent neural network model, and the activation value of the input layer is calculated by a Sigmoid activation function according to the alarm rules corresponding to the current tenant and the alarm rules corresponding to the current city. The specific formula is:

[0017] T l =S(w l *[I t-1 ,x t ]+b l )

[0018] Among them, T l represents the activation value of the input layer, S() represents the Sigmoid activation function, and w l represents the weight matrix, I t-1 Indicates the alarm rules corresponding to the current city, x t Indicates the alarm rule corresponding to the current tenant, b l Represents the bias term. The output layer calculates the activation value of the output layer through the Sigmoid activation function according to the cell states of the alarm rules corresponding to the current tenant, the alarm rules corresponding to the current city, and the alarm rules corresponding to the current city. The specific formula is:

[0019] T o =S(s o *[I t-1 ,x t ]+b o )

[0020] Among them, T o represents the activation value of the output layer, S() represents the Sigmoid activation function, and w o represents the weight matrix, It-1 Indicates the alarm rules corresponding to the current city, x t Indicates the alarm rule corresponding to the current tenant, b o Represents the bias term, which outputs the optimal estimated range (P1, P2) of the alarm rule corresponding to the current tenant through the fully connected layer.

[0021] Preferably, the floating values ​​corresponding to different time scales for different tenants are calculated, and the specific formula is:

[0022]

[0023] Among them, F st Indicates the floating value of the current tenant t corresponding to the previous m time scales, T t represents the traffic of the current tenant t, T s It represents the traffic of tenant t in the previous s time scale. The tenant weight is calculated according to the proportion of tenant traffic in the corresponding city of different tenants to the total traffic. The specific formula is:

[0024]

[0025] Among them, W t represents the tenant weight of the current tenant t in the corresponding city, T t It represents the total traffic of the current tenant t in the corresponding city, and calculates the floating value in different city dimensions by combining the average of the tenant floating value in different time scales. The specific formula is:

[0026]

[0027] Among them, F represents the floating value under different city dimensions, and the threshold corresponding to different alarm rules is determined to be (P1*F, P2*F).

[0028] Preferably, the alarm rules are divided into warning, serious, and emergency levels and correspond to low, medium, and high priorities. The alarm level, priority, and request format of the alarm rules of the Webhook interface are defined. Email, SMS, instant messaging tools, and telephone are supported as alarm notification channels. A notification callback mechanism is established to record the results of notification sending. Notification services are integrated and high-concurrency alarm scenarios are simulated.

[0029] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0030] Through lightweight data collectors, key business and system performance indicators are monitored in real time, and decision tree models are used to effectively screen features and detect anomalies in data to improve data quality. By integrating multiple data sources and implementing centralized log management, comprehensive analysis of car rental applications is ensured. At the same time, data is effectively cleaned and filled to improve statistical accuracy, and precise management and optimization of key business indicators and system performance indicators are achieved. This not only reduces data redundancy and errors, but also improves data integrity and consistency through effective detection and filling mechanisms.

[0031] By designing appropriate database table structures and indexes, efficient grouping and summarization of key business indicators and system performance indicators can be achieved. The GROUP BY aggregate function is used to improve data processing efficiency. At the same time, the ETL process is combined with batch writing to ensure regular data updates and efficient storage. The structured data storage method not only improves the speed of data retrieval and analysis, but also ensures the scalability and flexibility of the system.

[0032] By designing a flexible and powerful alarm rule configuration database and API interface, the alarm rules can be efficiently managed and updated, so as to monitor key business indicators and system performance indicators in real time. By setting a clear alarm trigger mechanism, the system can respond dynamically based on actual data, set reasonable thresholds for different indicators in a timely manner, and ensure rapid early warning and handling of potential problems. It not only improves the reliability of the service and user experience, but also effectively supports online car-hailing business decision-making, maintains continuous monitoring of the system health through regular inspections and notifications, and promotes the stable development and optimization of the business.

[0033] By calculating the floating values ​​and weights of tenants at different time scales, the alarm threshold can be adjusted according to real-time traffic changes, avoiding false alarms or missed alarms that may be caused by fixed thresholds, thereby improving the effectiveness of alarms. By analyzing the proportion of traffic of different tenants in a specific city, it is possible to better understand and characterize the relative importance of each tenant, thereby optimizing resource allocation and alarm strategies based on actual traffic distribution. By integrating multi-dimensional data, it ensures that alarm rules can be adapted to different scenarios, thereby enhancing the comprehensiveness and practicality of alarm rules.

[0034] By dividing the alarm rules into different levels such as warning, serious, and emergency, the importance and urgency of alarms can be effectively distinguished, ensuring that high-priority alarms can receive timely attention, while low-priority alarms can be appropriately delayed, optimizing the use of resources, supporting multiple notification channels such as email, SMS, instant messaging tools, and telephone, ensuring that relevant personnel can receive alarm notifications in a timely manner under any circumstances, improving the coverage and response speed of notifications, and establishing a notification callback mechanism to record the sending results in real time, ensure whether the message is delivered and effectively track the fault situation, define the alarm level, priority, and request format of the Webhook interface, and achieve standardization and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0036] Figure 1 The present invention is a method flow chart of a multi-tenant-based adaptive service alarm system.

[0037] Figure 2 It is a module schematic diagram of the multi-tenant-based adaptive service alarm system of the present invention. DETAILED DESCRIPTION

[0038] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0039] The present invention provides Figure 1 The multi-tenant-based adaptive business alarm system shown includes a data collection module, an aggregation storage module, an alarm rule module, an alarm calculation module, and an alarm notification module;

[0040] Data collection module: Set up a lightweight data collector to collect key business indicators and system performance indicators in real time, use the decision tree model to perform effective feature screening and anomaly detection, and perform cleaning and mean filling;

[0041] Select car rental application logs, car rental server performance, database, and network traffic as data sources, record the operation logs of the car rental application through centralized log management tools, set up lightweight data collectors to collect key business indicators and system performance indicators in real time, where key business indicators include but are not limited to the number of user registrations, customer retention rate, and average order value, where system performance indicators include but are not limited to response time, throughput, and error rate. Use the RESTful API interface to interact with the car rental application logs, use JDBC connections to interact with the car rental server performance and database, and use network monitoring tools to analyze and interact with network traffic.

[0042] Construct a tree structure, and let the branch nodes inside the tree structure represent the invalid features of key business indicators and system performance indicators, and the leaf nodes inside the tree structure represent the invalid removal results. Construct multiple decision trees by randomly selecting invalid features, key business indicators and system performance indicators, and randomly select invalid features of branch nodes according to leaf nodes for optimal splitting. Make independent classification predictions for each decision tree, and select the classification predictions that appear most frequently in all decision trees as invalid classifications. Judge and remove invalid key business indicators and system performance indicators, check and delete duplicate key business indicators and system performance indicators, check whether there are null values ​​and missing values ​​in key business indicators and system performance indicators, check whether key business indicators and system performance indicators are within the valid range, and check whether there are abnormalities and duplicate records in key business indicators and system performance indicators through uniqueness and consistency constraints. Determine the topic for different existence situations and divide different topic partitions, delete the corresponding log data objects through each partition, and use the mean filling method to fill the key business indicators and system performance indicators accordingly.

[0043] Aggregation storage module: Design a database table structure containing tenant ID and city index through MySQL, use GROUP BY aggregation function to group and summarize key business indicators and system performance indicators, implement ETL process and batch writing, and regularly store processing results;

[0044] According to the data source, the key business indicators and system performance indicators are grouped according to the tenants and cities. The GROUP BY aggregation function is used to aggregate the key business indicators and system performance indicators. MySQL is selected as the storage system. The database table structure is designed to include the fields of key business indicators and system performance indicators. Indexes are created with tenant ID and city. The ETL process is designed and the aggregation of key business indicators and system performance indicators is stored regularly through batch writing.

[0045] Alarm rule module: Design alarm rule configuration database table through MySQL, provide API interface to submit and update alarm rules, and set alarm trigger mechanism for response time, throughput, error rate, number of user registrations, customer retention rate and average order value;

[0046] Define alarm rule parameters including statistical range, calculation type and duration, select MySQL as the storage system, design the database table structure including the fields for alarm rule configuration, provide an API interface to allow submission of alarm rule configuration, provide an update interface to allow modification of existing alarm rules, support query of alarm rules by tenant ID through the API interface, use the background service to set the alarm trigger mechanism, and use the message queue to periodically check whether the key business indicators and system performance indicators of each tenant meet the alarm rules.

[0047] Define alarm rules, where the response time alarm rule counts the average response time within the last 5 minutes and sets the threshold to 1s. When the response time is greater than the threshold, an alarm is triggered. The throughput alarm rule counts the average throughput within the last 60 minutes and sets the threshold to less than 50 requests / second. When the throughput is greater than the threshold, an alarm is triggered. The error rate alarm rule counts the average value within the last 10 minutes and sets the threshold to 5%. When the error rate is greater than the threshold, an alarm is triggered. The user registration number alarm rule counts the daily registration number of user registrations within the last 24 hours and sets the threshold to 50% of the average on the same day of the previous week. When the user registration number is lower than the threshold, an alarm is triggered. The customer retention rate alarm rule counts the proportion of customers who continue to use the car rental service within a certain period of time within the last 30 days and sets the threshold to 70%. When the customer retention rate is lower than 70%, an alarm is triggered. The average order value alarm rule counts the average value within the last 30 days and sets the threshold to 10% of the average on the same day of the previous week. When the average order value is lower than the threshold, an alarm is triggered.

[0048] Alarm calculation module: Build a recurrent neural network model based on LSTM units, output the best estimated range of the alarm rules corresponding to the current tenant and city, and determine the tenant weight and floating value based on the traffic data at different time scales to determine the threshold of the corresponding alarm rule;

[0049] Select LSTM unit to build a recurrent neural network model. According to the alarm rules corresponding to the current tenant and the alarm rules corresponding to the current city, the activation value of the input layer is calculated through the Sigmoid activation function. The specific formula is:

[0050] T l =S(w l *[I t-1 ,x t ]+b l )

[0051] Among them, T l represents the activation value of the input layer, S() represents the Sigmoid activation function, and w l represents the weight matrix, I t-1 Indicates the alarm rules corresponding to the current city, x t Indicates the alarm rule corresponding to the current tenant, b l Represents the bias term. The output layer calculates the activation value of the output layer through the Sigmoid activation function according to the cell states of the alarm rules corresponding to the current tenant, the alarm rules corresponding to the current city, and the alarm rules corresponding to the current city. The specific formula is:

[0052] T o =S(s o *[I t-1 ,x t ]+b o )

[0053] Among them, T o represents the activation value of the output layer, S() represents the Sigmoid activation function, and w o represents the weight matrix, I t-1 Indicates the alarm rules corresponding to the current city, x t Indicates the alarm rule corresponding to the current tenant, b o Represents the bias term, which outputs the optimal estimated range (P1, P2) of the alarm rule corresponding to the current tenant through the fully connected layer.

[0054] The specific formula for calculating the floating values ​​of different tenants at different time scales is as follows:

[0055]

[0056] Among them, F st Indicates the floating value of the current tenant t corresponding to the previous m time scales, T t represents the traffic of the current tenant t, T s It represents the traffic of tenant t in the previous s time scale. The tenant weight is calculated according to the proportion of tenant traffic in the corresponding city of different tenants to the total traffic. The specific formula is:

[0057]

[0058] Among them, W t represents the tenant weight of the current tenant t in the corresponding city, T t It represents the total traffic of the current tenant t in the corresponding city, and calculates the floating value in different city dimensions by combining the average of the tenant floating value in different time scales. The specific formula is:

[0059]

[0060] Among them, F represents the floating value under different city dimensions, and the threshold corresponding to different alarm rules is determined to be (P1*F, P2*F).

[0061] Alarm notification module: By dividing alarm rules into levels and setting corresponding priorities, it supports multi-channel notifications and establishes a notification callback mechanism to record the sending results;

[0062] The alarm rules are divided into warning, serious, and emergency levels with corresponding low, medium, and high priorities. The alarm level, priority, and request format of the alarm rules of the Webhook interface are defined. Email, SMS, instant messaging tools, and phone are supported to implement alarm notification channels. A notification callback mechanism is established to record the results of notification sending. The notification service is integrated and high-concurrency alarm scenarios are simulated.

[0063] Through lightweight data collectors, key business and system performance indicators are monitored in real time, and decision tree models are used to effectively screen features and detect anomalies in data to improve data quality. By integrating multiple data sources and implementing centralized log management, comprehensive analysis of car rental applications is ensured. At the same time, data is effectively cleaned and filled to improve statistical accuracy, and precise management and optimization of key business indicators and system performance indicators are achieved. This not only reduces data redundancy and errors, but also improves data integrity and consistency through effective detection and filling mechanisms.

[0064] By designing appropriate database table structures and indexes, efficient grouping and summarization of key business indicators and system performance indicators can be achieved. The GROUP BY aggregate function is used to improve data processing efficiency. At the same time, the ETL process is combined with batch writing to ensure regular data updates and efficient storage. The structured data storage method not only improves the speed of data retrieval and analysis, but also ensures the scalability and flexibility of the system.

[0065] By designing a flexible and powerful alarm rule configuration database and API interface, the alarm rules can be efficiently managed and updated, so as to monitor key business indicators and system performance indicators in real time. By setting a clear alarm trigger mechanism, the system can respond dynamically based on actual data, set reasonable thresholds for different indicators in a timely manner, and ensure rapid early warning and handling of potential problems. It not only improves the reliability of the service and user experience, but also effectively supports online car-hailing business decision-making, maintains continuous monitoring of the system health through regular inspections and notifications, and promotes the stable development and optimization of the business.

[0066] By calculating the floating values ​​and weights of tenants at different time scales, the alarm threshold can be adjusted according to real-time traffic changes, avoiding false alarms or missed alarms that may be caused by fixed thresholds, thereby improving the effectiveness of alarms. Using LSTM networks to process historical data can more accurately predict future traffic changes based on past trends and patterns, making alarm rules more intelligent. By analyzing the proportion of traffic of different tenants in a specific city, it is possible to better understand and characterize the relative importance of each tenant, thereby optimizing resource allocation and alarm strategies based on actual traffic distribution. By integrating multi-dimensional data, it is ensured that alarm rules can be adapted to different scenarios, enhancing the comprehensiveness and practicality of alarm rules.

[0067] By dividing the alarm rules into different levels such as warning, serious, and emergency, the importance and urgency of alarms can be effectively distinguished, ensuring that high-priority alarms can receive timely attention, while low-priority alarms can be appropriately delayed, optimizing the use of resources, supporting multiple notification channels such as email, SMS, instant messaging tools, and telephone, ensuring that relevant personnel can receive alarm notifications in a timely manner under any circumstances, improving the coverage and response speed of notifications, and establishing a notification callback mechanism to record the sending results in real time, ensure whether the message is delivered and effectively track the fault situation, define the alarm level, priority, and request format of the Webhook interface, and achieve standardization and scalability.

[0068] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. The multi-tenant-based adaptive business alarm system is characterized by: It includes data collection module, aggregation storage module, alarm rule module, alarm calculation module and alarm notification module; Data collection module: Set up a lightweight data collector to collect key business indicators and system performance indicators in real time, use the decision tree model to perform effective feature screening and anomaly detection, and perform cleaning and mean filling; Aggregation storage module: Design a database table structure containing tenant ID and city index through MySQL, use GROUP BY aggregation function to group and summarize key business indicators and system performance indicators, implement ETL process and batch writing, and regularly store processing results; Alarm rule module: Design alarm rule configuration database table through MySQL, provide API interface to submit and update alarm rules, and set alarm trigger mechanism for response time, throughput, error rate, number of user registrations, customer retention rate and average order value; Alarm calculation module: Build a recurrent neural network model based on LSTM units, output the best estimated range of the alarm rules corresponding to the current tenant and city, and determine the tenant weight and floating value based on the traffic data at different time scales to determine the threshold of the corresponding alarm rule; Alarm notification module: By dividing alarm rules into levels and setting corresponding priorities, it supports multi-channel notifications and establishes a notification callback mechanism to record the sending results.

2. The multi-tenant-based adaptive service alarm system according to claim 1, characterized in that: In the data collection module, the car rental application log, car rental server performance, database, and network traffic are selected as data sources, the operation log of the car rental application is recorded through a centralized log management tool, a lightweight data collector is set to collect key business indicators and system performance indicators in real time, a tree structure is constructed to determine and remove invalid key business indicators and system performance indicators, duplicate key business indicators and system performance indicators are checked and deleted, key business indicators and system performance indicators are checked for null values ​​and missing values, key business indicators and system performance indicators are checked for being within a valid range, and key business indicators and system performance indicators are checked for abnormalities and duplicate records through uniqueness and consistency constraints, topics are determined for different existing situations and divided into different topic partitions, corresponding log data objects are deleted through each partition, and the key business indicators and system performance indicators are filled in accordingly using the mean filling method.

3. The multi-tenant-based adaptive service alarm system according to claim 2, characterized in that: The specific steps of constructing a tree structure to judge and remove invalid key business indicators and system performance indicators are as follows: the branch nodes inside the tree structure represent the invalid features of the key business indicators and system performance indicators, the leaf nodes inside the tree structure represent the invalid removal results, and multiple decision trees are constructed by randomly selecting invalid features, key business indicators and system performance indicators. The invalid features of the branch nodes are randomly selected according to the leaf nodes for optimal splitting, and each decision tree is independently classified and predicted, and the classification prediction that appears most frequently in all decision trees is selected as the invalid classification.

4. The multi-tenant-based adaptive service alarm system according to claim 1, characterized in that: In the aggregation storage module, the key business indicators and system performance indicators are grouped according to the tenants and cities according to the data source, and the aggregation of key business indicators and system performance indicators is realized by using the GROUP BY aggregation function. MySQL is selected as the storage system, and the database table structure is designed to include the fields of key business indicators and system performance indicators. Indexes are created with tenant ID and city. The aggregation of key business indicators and system performance indicators is stored regularly through batch writing and designing ETL process.

5. The multi-tenant-based adaptive service alarm system according to claim 1, characterized in that: In the alarm rule module, alarm rule parameters and alarm rules are defined, MySQL is selected as the storage system, the database table structure is designed including fields for alarm rule configuration, an API interface is provided to allow submission of alarm rule configuration, an update interface is provided to allow modification of existing alarm rules, the API interface supports querying alarm rules according to tenant ID, the background service is used to set the alarm trigger mechanism, and the message queue is used to periodically check whether the key business indicators and system performance indicators corresponding to each tenant meet the alarm rules.

6. The multi-tenant-based adaptive service alarm system according to claim 1, characterized in that: In the alarm calculation module, an LSTM unit is selected to construct a recurrent neural network model, and the activation value of the input layer is calculated by the Sigmoid activation function according to the alarm rules corresponding to the current tenant and the alarm rules corresponding to the current city. The output layer calculates the activation value of the output layer by the Sigmoid activation function according to the alarm rules corresponding to the current tenant and the alarm rules corresponding to the current city and the cell state of the alarm rules corresponding to the current city. The optimal estimated range (P1, P2) of the alarm rules corresponding to the current tenant is output through the fully connected layer, and the floating values ​​corresponding to different time scales for different tenants are calculated. The tenant weight is calculated according to the proportion of tenant traffic in different tenant corresponding cities to the total traffic, and the floating values ​​in different city dimensions are calculated in combination with the average of the tenant floating values ​​under different time scales, and the thresholds corresponding to different alarm rules are finally determined to be (P1*F, P2*F).

7. The multi-tenant-based adaptive service alarm system according to claim 6, characterized in that: The specific formula for calculating the activation value of the input layer is: T l =S(w l *[I t-1 ,x t ]+b l ) Among them, T l represents the activation value of the input layer, S() represents the Sigmoid activation function, and w l represents the weight matrix, I t-1 Indicates the alarm rules corresponding to the current city, x t Indicates the alarm rule corresponding to the current tenant, b l Represents the bias term, and the specific formula for calculating the activation value of the output layer is: T o =S(s o *[I t-1 ,x t ]+b o ) Among them, T o represents the activation value of the output layer, S() represents the Sigmoid activation function, and w o represents the weight matrix, I t-1 Indicates the alarm rules corresponding to the current city, x t Indicates the alarm rule corresponding to the current tenant, b o represents the bias term, and the specific formula for the floating values ​​of different tenants corresponding to different time scales is: Among them, F st Indicates the floating value of the current tenant t corresponding to the previous m time scales, T t represents the traffic of the current tenant t, T s represents the traffic of tenant t in the previous s time scale. The specific formula for calculating the tenant weight is: Among them, W t represents the tenant weight of the current tenant t in the corresponding city, T t It represents the total traffic in the city corresponding to the current tenant t. The specific formula for calculating the floating value in different city dimensions is: Among them, F represents the floating value under different city dimensions.

8. The multi-tenant-based adaptive service alarm system according to claim 1, characterized in that: In the alarm notification module, the alarm rules are divided into warning, serious, and emergency levels and correspond to low, medium, and high priorities. The alarm level, priority, and request format of the alarm rules of the Webhook interface are defined. Email, SMS, instant messaging tools, and telephone are supported to implement alarm notification channels. A notification callback mechanism is established to record the notification sending results. The notification service is integrated and a high-concurrency alarm scenario is simulated.

Images