General framework for monitoring health indexes of meteorological service system

Abstract

The invention discloses a general framework for monitoring health indexes of a meteorological service system, and belongs to the technical field of comprehensive monitoring of meteorological data. Comprising an acquisition layer, a storage layer, a calculation layer and a service layer. The acquisition layer adopts an REST interface and a probe script to realize access of health indexes of a third-party meteorological service system; the storage layer adopts different storage models to store monitoring data into corresponding databases according to different characteristics of the monitoring data; the calculation layer realizes convergence calculation and alarm of health indexes of the meteorological service system; and the service layer realizes visual display of product integrity, fault tracing, cloud resources and an application end of the meteorological service system. According to the invention, health index management of all meteorological service systems can be realized based on configuration, when a certain health index is abnormal, a fault link can be rapidly positioned, early warning information can be accurately pushed, the troubleshooting efficiency is significantly improved, and a guarantee is provided for the operation stability of the meteorological service systems.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of meteorological data comprehensive monitoring, in particular to a meteorological business system health index monitoring general framework. BACKGROUND

[0002] High-quality promotion of meteorological modernization construction, and improvement of the quality and efficiency of meteorological modernization construction are the development direction of meteorological business. At present, the main problem faced by meteorological business system operation and maintenance is that alarm fault positioning still needs manual investigation, and due to many system flow links and many involved personnel, etc., it leads to long time consumption in analyzing and processing fault causes, and cannot achieve effective correlation between different meteorological business system alarms. Therefore, it is urgent to design a general meteorological business system health index evaluation framework, form general evaluation indexes for attention to system stability and other aspects, and realize the operation and maintenance change from passive alarm to active early warning. At the same time, with the online of meteorological big data cloud platform business, more meteorological business systems are integrated into the cloud platform, and alarm information is put into the "Tianjing". The existing fault positioning is usually carried out after the alarm is issued, and the problems of many manual operation and maintenance positioning links and long time consumption are highlighted. In order to better utilize and expand the monitoring capability of the "Tianjing" system, it is necessary to increase the research on the evaluation capability and early warning mechanism of the core business system health index, so as to further improve the stability of the business system. SUMMARY

[0003] In order to solve the defects of the prior art, the present application provides a meteorological business system health index monitoring general framework, which can realize the management of all meteorological business system health indexes based on configuration, can quickly locate the fault link when a certain health index is abnormal, and can accurately push the early warning information, thereby significantly improving the fault troubleshooting efficiency and providing protection for the stability of the meteorological business system operation.

[0004] The technical scheme provided by the present application is as follows:

[0005] A meteorological business system health index monitoring general framework, comprising a collection layer, a storage layer, a calculation layer and a service layer, wherein:

[0006] The collection layer realizes the access of the health indexes of the third-party meteorological business system by using REST interface and probe script;

[0007] The storage layer stores the monitoring data into the corresponding database by using different storage models according to different characteristics of the monitoring data;

[0008] The calculation layer realizes the aggregation calculation and alarm of the health indexes of the meteorological business system;

[0009] The service layer realizes the visual display of the product integrity, fault tracing, cloud resources and application end of the meteorological business system.

[0010] Further, in the storage layer, the meteorological business system health indicators are stored in an ElasticSearch database based on index technology, the alarms and configuration parameters are stored in a MongoDB database, and the hotspot data is stored in a Redis in-memory database.

[0011] Further, the computing tasks of the computing layer are based on the threshold values of the meteorological business system health indicators in the MongoDB database, the business time of the meteorological business system health indicators is sorted, and the normal operation range of the meteorological business system health indicators is determined through a statistical algorithm. If abnormal, the business system health is evaluated according to the business configuration weight rules, and an alarm is generated and targeted pushed.

[0012] Further, the types of the meteorological business system health indicators include business capability health indicators, business system performance health indicators, and business system availability indicators.

[0013] Further, the business capability health indicators are calculated by the following method:

[0014] The business system sends the current business volume obtained through monitoring to the Tianjing according to the set statistical period and format, or the Tianjing's collection agent program collects the business system's logs and analyzes and calculates the current business volume;

[0015] The Tianjing calculates the business capability health indicators according to the obtained current business volume and the maximum business carrying capacity of the business system;

[0016] The business capability health indicators = current business volume / maximum business carrying capacity.

[0017] Further, the business system performance health indicators are calculated by the following method:

[0018] The business system sends the business system performance running value obtained through monitoring to the Tianjing according to the set statistical period and format, or the Tianjing's collection agent program collects the business system's logs and analyzes and calculates the business system performance running value;

[0019] The Tianjing calculates the business system performance health indicators according to the obtained business system performance running value and the business system performance threshold value;

[0020] The business system performance health indicators = business system performance running value / business system performance threshold value.

[0021] Further, the method further comprises:

[0022] According to the waveforms of the business capability health indicators and the business system performance health indicators, corresponding algorithms are allocated for abnormal monitoring.

[0023] Further, for the business capability health indicator, an initial value of a health indicator threshold of the business capability health indicator is set, and an indicator health degree is calculated through an actual running value of the business capability health indicator of the current meteorological business system / health indicator threshold;

[0024] For the business capability health indicator, a 3-sigma algorithm is assigned to dynamically adjust the health indicator threshold, and for the case of sudden increase or sudden decrease of the indicator, an inflection point of time series data is calculated, abnormal monitoring is performed, and an alarm is issued;

[0025] If the indicator health degree is in the interval of 0 to 20%, it is considered that the business capability health indicator is in a healthy state, and no alarm is given; if the indicator health degree is in the interval of 20%-80%, it is considered that the business capability health indicator is in a sub-healthy state, general alarm information is issued to notify the system administrator to pay attention to the running state of the system, and general alarm information is sent once every 10% increase of the indicator health degree; if the indicator health degree exceeds 80%, it is considered that the business capability health indicator is in a serious alarm state, and serious alarm information is issued to notify the system administrator to handle the fault.

[0026] Further, for the business system performance health indicator, an initial value of a health indicator threshold of the business system performance health indicator is set, and an indicator health degree is calculated through an actual running value of the business system performance health indicator of the current meteorological business system / health indicator threshold;

[0027] For the business system performance health indicator, an isolated forest algorithm is assigned to mark abnormal discrete data of the indicator, and an alarm is issued.

[0028] Further, the business capability health indicator and the business system performance health indicator both include three dimensions of a large category, a small category, and an indicator, a bottom-up weight aggregation model is performed from the judgment of the indicator dimension, the score of the small category dimension is calculated according to the weight of each indicator, the score of the small category dimension is calculated according to the weight of the large category dimension, and the score of the large category dimension is calculated according to the weight of the entire business capability health indicator / business system performance health indicator;

[0029] And in the indicator dimension, different alarm levels are reduced by weight, normal without reduction, general alarm level reduced by 50%, and serious alarm level reduced by 100%, so as to reduce the authority of the corresponding subcategory, and then the weight of the large category is reduced according to the weight ratio of the subcategory. Each category is calculated according to the different weight ratio.

[0030] Further, the business capability health indicator and the business system performance health indicator are respectively predicted and warned by a quadratic exponential smoothing algorithm and an LSTM algorithm for the trend in a future set time period.

[0031] Further, the business system availability index is calculated by the following method:

[0032] The business system running time is recorded to obtain the average failure-free time, the interval time between two start-stop is recorded to obtain the average failure recovery time, and the business system availability index is calculated;

[0033] The business system availability index = average failure-free time / (average failure-free time + average failure recovery time).

[0034] The present application has the following beneficial effects:

[0035] The present application classifies the meteorological business system health index on one hand, can directly obtain the abnormal information of the meteorological business system running time, mainly including core business system algorithm scheduling, interface service, key data storage time-consuming, whether the storage service is normal, etc. On the other hand, the system availability evaluation based on the business system failure time is established, which can realize the management of all meteorological business system health indexes based on the configuration. When a health index is abnormal, the threshold can be matched automatically, the fault link can be quickly located, and the warning information can be accurately pushed, which can significantly improve the fault troubleshooting efficiency and provide protection for the stability of the meteorological business system running. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 It is a structure diagram of the meteorological business system health index monitoring universal framework of the present application;

[0037] Figure 2 It is an example of the business capability health index and its weight configuration;

[0038] Figure 3 It is a comparison of the prediction results of the residual Seq2Seq-LSTM model, the Transformer model, the ES exponential smoothing method and the Arima model on the concurrent access peak (times) index. DETAILED DESCRIPTION

[0039] In order to make the technical problems, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in conjunction with the drawings and specific embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application generally described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0040] The embodiments of the present application provide a general framework for monitoring health indicators of meteorological business systems, as shown in Figure 1 The general framework comprises a collection layer, a storage layer, a calculation layer and a service layer, wherein:

[0041] The collection layer realizes efficient access to health indicators of third-party meteorological business systems by using REST interface and probe scripts.

[0042] The storage layer stores the monitoring data into corresponding databases by using different storage models according to different characteristics of the monitoring data.

[0043] Among them, the health indicators of meteorological business systems are stored in an ElasticSearch database based on index technology, the alarms and configuration parameters are stored in a MongoDB database, and the hot data is stored in a Redis in-memory database.

[0044] The calculation layer is the core layer of the entire architecture, which realizes the aggregation calculation and alarm of the health indicators of meteorological business systems.

[0045] Specifically, the calculation task sorts the business time of the health indicators of meteorological business systems based on the threshold values of the health indicators of meteorological business systems configured in the MongoDB database, and determines the normal operation range of the health indicators of meteorological business systems by using statistical algorithms. If it is abnormal, the health of the business system is evaluated according to the weight rules of the business, and an alarm is generated and targeted pushed.

[0046] The service layer realizes the visual display of the integrity of meteorological business system products, fault tracing, cloud resources and application end.

[0047] The types of the aforementioned health indicators of meteorological business systems include business capability health indicators, business system performance health indicators and business system availability indicators.

[0048] The business capability health index (HEALTH) reflects the current business capability that the business system can support. In one example, the business capability health index includes three dimensions of major category, minor category, and index, as shown in Table 1 below.

[0049] Table 1: Business capability health index

[0050]

[0051] Two ways are supported to obtain the index content. One way is that the business system sends the current business volume obtained by monitoring to the Tianjing according to the set statistical period and format. Another way is that the Tianjing obtains the current business volume by analyzing and calculating the logs of the business system collected by the collection agent program.

[0052] Then, the Tianjing obtains the business capability health index by calculating the current business volume and the maximum business carrying capacity of the business system.

[0053] The business capability health index = current business volume / maximum business carrying capacity.

[0054] The business system performance health index (PREF) reflects the current performance state of the business system. In one example, the business system performance health index includes three dimensions of major category, minor category, and index, as shown in Table 2 below.

[0055] Table 2: Business system performance health index

[0056]

[0057] Two ways are supported to obtain the index content. One way is that the business system sends the business system performance running value obtained by monitoring to the Tianjing according to the set statistical period and format. Another way is that the Tianjing obtains the business system performance running value by analyzing and calculating the logs of the business system collected by the collection agent program.

[0058] Then, the Tianjing obtains the business system performance health index by calculating the business system performance running value and the business system performance threshold.

[0059] The business system performance health index = business system performance running value / business system performance threshold.

[0060] After the business capability health index and the business system performance health index are calculated, corresponding algorithms can be allocated according to the waveforms of the business capability health index and the business system performance health index for abnormal monitoring.

[0061] For the business capability health index, set the initial value of the health index threshold of the business capability health index, reflecting the maximum carrying capacity of the current system business. And calculate the index health degree through the actual running value / health index threshold of the business capability health index of the current meteorological business system.

[0062] For the business capability health index, the time series generated by the stable operation of this type of health index is usually a stable waveform, and the oscillation amplitude is within a fixed range. Therefore, according to the waveform of the index, 3-sigma algorithm is allocated to dynamically adjust the health index threshold, and the inflection point of the time series data is calculated for the case of sudden increase or sudden decrease of the index, to perform abnormal monitoring and issue an alarm;

[0063] Among them, if the index health degree is in the interval of 0 to 20%, the business capability health index is considered to be in a healthy state, and no alarm is given; if the index health degree is in the interval of 20%-80%, the business capability health index is considered to be in a sub-healthy state, and general alarm information is issued to notify the system administrator to pay attention to the running state of the system, and as the index health degree increases, general alarm information is sent once every 10%; if the index health degree exceeds 80%, the business capability health index is considered to be in a serious alarm state, and serious alarm information is issued to notify the system administrator to handle the fault.

[0064] For the business system performance health index, the initial value of the health index threshold of the business system performance health index also needs to be set, reflecting the performance state of the current system business. And calculate the index health degree through the actual running value / health index threshold of the business system performance health index of the current meteorological business system.

[0065] For the business system performance health index, the time series generated by the stable operation of this type of health index is usually a large waveform with disordered oscillation amplitude. Therefore, according to the waveform of the index, isolated forest algorithm is allocated to mark the abnormal discrete data of the index, and an alarm is issued.

[0066] The present application uses appropriate algorithms for abnormal monitoring of business capability health index and business system performance health index according to the characteristics of their data waveforms, and achieves good monitoring effect. The comparison of the two algorithms is shown in Table 3.

[0067] Table 3: Comparison of 3-sigma method and isolated forest algorithm

[0068]

[0069] When calculating business capability health indicators / business system performance health indicators, the calculation is performed according to three dimensions (i.e., major category, minor category, and indicator). A bottom-up weighted aggregation model is used based on the indicator dimension judgment. Each indicator is weighted to calculate the score of the minor category dimension (if no weight is configured, it is calculated by equal weight; if weight is configured, it is calculated by the configured weight). The scores of the minor category dimensions are weighted to calculate the score of the major category dimension. The scores of the major category dimensions are weighted to calculate the score of the entire business capability health indicator / business system performance health indicator. Figure 2 This is an example of business capability health indicators and their weight configuration.

[0070] Furthermore, in terms of metrics, different alarm levels are weighted and reduced. Normal alarms are not reduced, general alarms are reduced by 50%, and severe alarms are reduced by 100%. This reduces the permissions of the corresponding subclasses. Then, the weight of the major category is reduced according to the weight ratio of the subclasses. Each category is calculated and reduced according to its corresponding weight ratio to calculate the specific value of the current health metric.

[0071] Continuous monitoring and calculation are performed, and the health index value of the business system is calculated by summarizing the alarm information of the business system health indicators in the past hour.

[0072] This invention can also predict and warn about health indicators. Specifically, it uses a quadratic exponential smoothing algorithm and an LSTM algorithm to predict and warn about the trends of business capability health indicators and business system performance health indicators within a set future time period.

[0073] The prediction algorithms for business capability health indicators and business system performance health indicators are also determined based on their waveforms, tailored to their respective characteristics. Quadratic exponential smoothing, also known as Holt's linear trend method, can handle data without a clear trend (i.e., data fluctuates around a relatively stable level). Holt's method, by introducing a second smoothing equation, explicitly models and predicts linear trends in time series. Its core idea is that a time series not only has a level that changes over time, but also a trend / slope. This method uses two smoothing parameters to update estimates of the current level and current trend, respectively, and uses these estimates to make future predictions, aiming to more accurately predict time series with linear trends (upward or downward) but without obvious seasonality.

[0074] Long Short-Term Memory (LSTM) algorithm, designed to solve the problem of gradient vanishing / explosion in traditional RNN when processing long sequence data. Its core idea is to selectively retain important information and discard irrelevant information, thus performing well in time series data prediction (such as weather monitoring, equipment status analysis), and suitable for dynamic modeling of abnormal and unordered patterns in time series.

[0075] Figure 3 is the residual Seq2Seq-LSTM model, the Transformer model, the ES exponential smoothing method, and the Arima model on the prediction results of the concurrent access peak (times) index. Table 4 is the specific experimental index results of each model.

[0076] Table 4: Model test results

[0077]

[0078] From the experimental results, it can be seen that in the concurrent access peak (times) index, the residual Seq2Seq-LSTM model prediction value is relatively consistent with the true value, and can still maintain a relatively stable prediction; while the response ability of the Transformer model and the ES exponential smoothing method to the sharp fluctuation is obviously weaker than that of the residual Seq2Seq-LSTM model; the Arima model has a large prediction deviation in the early stage.

[0079] For the business system availability index, the calculation method is as follows:

[0080] Record the running time of the business system to obtain the average failure-free time, record the interval time between two starts and stops to obtain the average failure recovery time, and calculate the business system availability index.

[0081] Among them, the business system availability index = average failure-free time / (average failure-free time + average failure recovery time).

[0082] Specifically, when a business failure occurs, the associated query submits a business system failure work order, and the general monitoring framework can obtain the system interruption time, so that the annual system availability can be calculated as the basis for system evaluation.

[0083] The aspect of the present application formulates a meteorological business system health index classification specification, can directly obtain abnormal information of meteorological business system running, mainly including core business system algorithm scheduling, interface service, key data storage time, whether storage service is normal and the like. The other aspect establishes a system availability evaluation based on business system fault time, can realize all meteorological business system health index management based on configuration, when a health index is abnormal, can automatically match threshold, quickly locate fault link, and accurately push early warning information, significantly improves fault troubleshooting efficiency, and provides guarantee for stability of meteorological business system operation.

[0084] Finally, it should be noted that: the above-described embodiments are only specific embodiments of the present application, for explaining the technical solutions of the present application, but not limiting them, the protection scope of the present application is not limited to this, although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand: any person skilled in the art in the technical range disclosed by the present application, still can modify or easily think of changes to the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A general framework for monitoring health indicators of meteorological operational systems, characterized in that, The system comprises a collection layer, a storage layer, a calculation layer and a service layer, wherein: The collection layer uses REST interface and probe scripts to realize the access of health indicators of third-party meteorological service systems; The storage layer uses different storage models to store the monitoring data into corresponding databases according to different characteristics of the monitoring data; The calculation layer realizes the aggregation calculation and alarm of health indicators of meteorological service systems; The service layer realizes the visual display of product integrity, fault tracing, cloud resources and application end of meteorological service systems.

2. The weather business system health indicator monitoring general framework of claim 1, wherein, In the storage layer, the health indicators of meteorological service systems are stored in an ElasticSearch database based on index technology, the alarms and configuration parameters are stored in a MongoDB database, and the hot data is stored in a Redis in-memory database.

3. The weather business system health indicator monitoring general framework of claim 2, wherein, The calculation task of the calculation layer is based on the threshold value of the health indicators of meteorological service systems in the MongoDB database, the business time of the health indicators of meteorological service systems is sorted, and the normal operation range of the health indicators of meteorological service systems is determined through statistical algorithms. If it is abnormal, the health of the business system is evaluated according to the weight rules of the business, and an alarm is generated and targeted pushed.

4. The general framework for monitoring the health of meteorological business systems according to any of claims 1 to 3, characterized in that, The types of the health indicators of meteorological service systems include business capacity health indicators, business system performance health indicators and business system availability indicators.

5. The weather business system health indicator monitoring general framework of claim 4, wherein, The business capacity health indicators are calculated by the following method: The business system sends the current business volume obtained by monitoring to Tianjing according to the set statistical period and format, or the collection agent program of Tianjing collects the logs of the business system and analyzes and calculates the current business volume; Tianjing calculates the business capacity health indicators according to the obtained current business volume and the maximum business carrying capacity of the business system; Wherein, the business capacity health indicators = current business volume / maximum business carrying capacity.

6. The weather business system health indicator monitoring general framework of claim 5, wherein, The business system performance health indicators are calculated by the following method: The business system sends the business system performance running value obtained by monitoring to Tianjing according to the set statistical period and format, or the collection agent program of Tianjing collects the logs of the business system and analyzes and calculates the business system performance running value; Tianjing calculates the business system performance health indicators according to the obtained business system performance running value and the business system performance threshold value; Wherein, the business system performance health indicators = business system performance running value / business system performance threshold value.

7. The weather business system health indicator monitoring general framework of claim 6, wherein, The method further comprises: According to the waveforms of the business capacity health indicators and the business system performance health indicators, corresponding algorithms are allocated for abnormal monitoring.

8. The weather business system health indicator monitoring general framework of claim 7, wherein, For the business capacity health indicators, the initial value of the health indicator threshold value of the business capacity health indicators is set, and the index health degree is calculated by the actual running value of the business capacity health indicators of the current meteorological service system / the health indicator threshold value; For the business capacity health indicators, the 3-sigma algorithm is allocated to dynamically adjust the health indicator threshold value, and the inflection point of the time series data is calculated for the case of sudden increase or decrease of the index, so as to perform abnormal monitoring and issue an alarm. If the index health degree is in the interval of 0-20%, the business capability health index is considered to be in a healthy state, and no alarm is given; if the index health degree is in the interval of 20%-80%, the business capability health index is considered to be in a sub-healthy state, a general alarm information is sent to notify the system administrator to pay attention to the running state of the system, and with the increase of the index health degree, a general alarm information is sent once every 10% increase; If the index health degree exceeds 80%, the business capability health index is considered to be in a serious alarm state, a serious alarm information is sent to notify the system administrator to perform fault processing.

9. The weather business system health indicator monitoring general framework of claim 8, wherein, For the business system performance health index, an initial value of a health index threshold of the business system performance health index is set, and an index health degree is calculated through an actual running value of the business system performance health index of the current meteorological business system / the health index threshold; For the business system performance health index, an isolated forest algorithm is allocated to mark abnormal discrete data of the index, and an alarm is given.

10. The weather business system health indicator monitoring general framework of claim 9, wherein, The business capability health index and the business system performance health index both include three dimensions of a large category, a small category and an index, a bottom-up weight aggregation model is performed from the judgment of the index dimension, each index is calculated to obtain a score of the small category dimension according to a weight, the score of the small category dimension is calculated to obtain a score of the large category dimension according to a weight, and the score of the large category dimension is calculated to obtain a score of the entire business capability health index / business system performance health index according to a weight; And in the index dimension, different alarm levels are reduced by weight, and no reduction is performed for normal, a reduction of 50% is performed for a general alarm level, a reduction of 100% is performed for a serious alarm level, and the corresponding subcategory is reduced in this way, and then the weight of the large category is reduced according to the weight ratio of the subcategory, and each category is calculated according to the corresponding different weight ratio.

11. The weather business system health indicator monitoring general framework of claim 9, wherein, The trends of the business capability health index and the business system performance health index in a future set time period are predicted and warned through a quadratic exponential smoothing algorithm and an LSTM algorithm respectively.

12. The weather business system health indicator monitoring general framework of claim 4, wherein, The business system availability index is calculated by the following method: The running duration of the business system is recorded to obtain the mean time between failures, the interval duration between two start-stop times is recorded to obtain the mean time to recover from failure, and the business system availability index is calculated; The business system availability index = mean time between failures / (mean time between failures+mean time to recover from failure).

Images