A terminal service performance monitoring optimization system and method based on data fusion

Abstract

The application discloses a kind of terminal service performance monitoring optimization system and method based on data fusion, it is related to terminal service performance monitoring optimization technical field, the method includes the following steps: obtaining relevant performance data;According to the intersection condition of actual receiving performance data set and the intersection condition of the preset necessary performance data set, define the performance data set completeness that terminal reports to cloud, utilize terminal and the terminal hardware resource consumption data that full volume terminal reports to cloud, obtain corresponding prediction effect by the pre-trained performance optimization effect prediction model, calculate prediction effect deviation parameter;According to the performance data set completeness that terminal reports to cloud in historical time period and corresponding moment prediction effect deviation, establish quantification model, set prediction effect deviation threshold value, when deviation does not exceed this threshold value, cloud issues optimization strategy to terminal;Strategy credibility factor is calculated, refer to strategy credibility factor threshold value and execute strategy action, the application realizes terminal service performance monitoring optimization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of terminal service performance monitoring optimization, and particularly relates to a terminal service performance monitoring optimization system and method based on data fusion. BACKGROUND

[0002] With the rapid development of Internet of Things and mobile Internet technology, the service scenarios of terminal devices are increasingly complex, and the monitoring and optimization of terminal service performance directly affect user experience and resource allocation efficiency, and become a core link of enterprise digital operation.

[0003] However, the traditional terminal service performance monitoring and optimization method often faces the following problems when dealing with multi-source heterogeneous data processing and dynamic strategy adjustment. First, the data acquisition and fusion capability is insufficient, and it is difficult to integrate full-dimensional performance data. Terminal service performance data involves multi-source heterogeneous data of hardware resources, application code execution path, network transmission, business process and cloud resources. The traditional method mainly uses a single data source, which cannot realize the deep fusion of heterogeneous data, resulting in poor data integrity and one-sided dimension, and hidden dangers for subsequent performance analysis. Second, the performance deviation evaluation lacks quantitative support, and the optimization effect is unstable. The traditional performance optimization relies on experience model prediction effect, but does not consider the influence of the completeness of terminal reported data on the prediction result, cannot quantify the correlation between data completeness and prediction deviation, and leads to large prediction result deviation and insufficient optimization strategy pertinence. Third, the strategy issuing lacks scientific threshold judgment and relies on manual decision. The traditional method mainly sets the execution condition based on subjective experience when issuing the optimization strategy, and does not establish a quantitative model to determine a reasonable deviation threshold through historical data. SUMMARY

[0004] The present application aims to provide a terminal service performance monitoring optimization system and method based on data fusion to solve the problems in the prior art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a terminal service performance monitoring optimization method based on data fusion, which comprises the following steps:

[0006] Obtain terminal hardware resource consumption data and application layer code execution path, and integrate heterogeneous data of network, business and cloud resources;

[0007] According to the intersection of the actual received performance data set and the preset necessary performance data set, define the completeness of the performance data set reported by the terminal to the cloud, use the performance data set reported by the terminal and the full amount of terminal reported data, obtain the corresponding prediction effect through the pre-trained performance optimization effect prediction model, and calculate the prediction effect deviation parameter;

[0008] According to the completeness of the performance data set reported by the terminal to the cloud and the prediction effect deviation at the corresponding moment in the historical time period, a quantitative model is established, and a prediction effect deviation threshold is set. When the deviation does not exceed the threshold, the cloud issues an optimization strategy to the terminal;

[0009] The calculation strategy credibility factor is used to execute the strategy action by referring to the strategy credibility factor threshold.

[0010] The terminal hardware resource consumption data and the application layer code execution path are obtained, and the heterogeneous data of network, service and cloud resources are integrated. The specific steps include:

[0011] The terminal hardware resource consumption data is collected by using the dynamic insertion technology, denoted as set H = {h1, h2,..., hm}, wherein h1, h2,..., hm represent the 1st, 2nd,..., mth terminal hardware resource consumption data, and m represents the number of terminal hardware resource consumption data types;

[0012] The application layer code execution path is parsed by combining the AST abstract syntax tree, denoted as code path set C = {c1, c2,..., cp}, wherein c1, c2,..., cp represent the 1st, 2nd,..., pth code execution path, and p represents the total number of code execution paths;

[0013] The network transmission, service process and cloud resource type data are integrated. The specific steps are as follows: the network transmission data is collected by using non-intrusive bypass monitoring technology, the service process data is collected by using distributed transaction tracking technology, and the cloud resource data is collected by using virtualization layer Hypervisor monitoring probe technology.

[0014] According to the intersection of the actual received performance data set and the preset necessary performance data set, the completeness of the performance data set reported by the terminal to the cloud is defined, and the corresponding prediction effect is obtained by using the performance data set reported by the terminal and the full amount of terminal reported data through the pre-trained performance optimization effect prediction model. The specific steps include:

[0015] The completeness of the performance data set reported by the terminal to the cloud at ti moment is calculated based on the actual received performance data set and the preset necessary performance data set at ti moment, and is defined as follows: ati = (S1 / S2)*100%;

[0016] Wherein ati represents the completeness of the performance data set reported by the terminal to the cloud at ti moment, S1 represents the number of elements in the intersection of the actual cloud received performance data set and the preset necessary performance data set; S2 represents the number of elements in the preset performance data set, and the performance data set includes terminal hardware resource consumption data, application layer code execution path, network, service and cloud resource heterogeneous data;

[0017] Based on the terminal hardware resource consumption data actually reported by the terminal to the cloud at time t, input into the pre-trained performance optimization effect prediction model, output the actual prediction optimization effect at time t, wherein the actual prediction optimization effect represents the percentage of reduction of the terminal hardware resource consumption data actually reported by the terminal to the cloud;

[0018] The terminal hardware resource consumption data reported by the terminal to the cloud at time t is input into the performance optimization effect prediction model, and the benchmark prediction effect at time t is output. The prediction effect deviation parameter at time t is calculated, which is defined as follows: b = |A1-A2| / A2*100%; wherein b represents the benchmark prediction effect, A1 represents the prediction optimization effect at time t, and A2 represents the benchmark prediction effect at time t. The terminal hardware resource consumption data reported by the terminal to the cloud represents the complete terminal hardware resource consumption data in an ideal state, and the benchmark prediction effect represents the percentage of reduction of the terminal hardware resource consumption data reported by the terminal to the cloud.

[0019] According to the completeness of the performance data set reported by the terminal to the cloud in the historical time period and the prediction effect deviation at the corresponding time, a quantitative model is established, and a prediction effect deviation threshold is set. When the deviation does not exceed the threshold, the cloud issues an optimization strategy to the terminal, and the specific steps include:

[0020] According to the completeness of the performance data set reported by the terminal to the cloud in the historical time period and the prediction effect deviation parameter at the corresponding time, a quantitative model of data completeness to prediction effect deviation parameter is established by least square regression, which is denoted as b = f(ati) + ε; wherein b represents the prediction effect deviation parameter, f(ati) represents the fitting function, and ε represents the random error term;

[0021] A prediction effect deviation parameter threshold is set, and the prediction effect deviation is compared with the set prediction effect deviation parameter threshold. When it does not exceed the set prediction effect deviation parameter threshold, the cloud issues an optimization strategy to the terminal.

[0022] The strategy credibility factor is calculated, and the strategy action is executed by referring to the strategy credibility factor threshold, and the specific steps include:

[0023] The strategy credibility factor is calculated, which is defined as follows: φ = 1-b; wherein φ represents the credibility factor;

[0024] For each historical moment, the confidence factor calculated at the corresponding moment and the policy execution result are recorded, wherein the policy execution result includes policy execution success and policy execution failure, the policy execution success is determined when the measured performance optimization effect reaches z times of the predicted performance optimization effect at ti moment, the policy execution failure is determined when the measured performance optimization effect does not reach z times of the predicted performance optimization effect, z represents the multiple, wherein the measured performance optimization effect represents the percentage of reduction of the terminal hardware resource consumption data actually observed after the terminal executes the optimization strategy issued by the cloud; the predicted performance optimization effect represents the average value of the baseline prediction effect at ti moment and the actual prediction optimization effect at ti moment;

[0025] The confidence factor is interval divided, the success rate of the policy in each interval is counted, the minimum acceptable success rate is determined according to the business demand, and the minimum confidence factor meeting the success rate is found as the execution threshold; when the confidence factor exceeds the execution threshold, the corresponding policy is executed, otherwise the policy is not executed.

[0026] A terminal service performance monitoring optimization system based on data fusion, the system comprises: a data acquisition module, a performance data analysis module, a quantitative modeling and policy issuing module and a policy evaluation module, the data acquisition module is used for acquiring terminal hardware resource consumption data and application layer code execution path, and integrating heterogeneous data of network, service and cloud resources; the performance data analysis module is used for acquiring corresponding prediction effect by using the performance data set reported by the terminal and the full terminal reported data, and calculating the prediction effect deviation parameter through the pre-trained performance optimization effect prediction model; the quantitative modeling and policy issuing module is used for establishing a quantitative model according to the performance data set reported by the terminal to the cloud in a historical time period and the prediction effect deviation at the corresponding moment, and setting a prediction effect deviation threshold value, when the deviation does not exceed the threshold value, the cloud issues an optimization strategy to the terminal; the policy evaluation module is used for calculating a policy confidence factor according to the performance optimization effect of the actual performance data set and the full performance data set at the corresponding moment, and executing the policy action by referring to the policy confidence factor threshold.

[0027] The data acquisition module comprises a terminal hardware resource acquisition unit, an application layer code path analysis unit and a heterogeneous data integration unit, the terminal hardware resource acquisition unit is used for acquiring terminal hardware resource consumption data by using dynamic insertion technology; the application layer code path analysis unit is used for analyzing the application layer code execution path in combination with the AST abstract syntax tree; the heterogeneous data integration unit is used for integrating network transmission, service process and cloud resource type data.

[0028] The performance data analysis module comprises a performance data completeness calculation unit, an optimization effect prediction unit and a predicted effect deviation parameter calculation unit, the performance data completeness calculation unit is used for calculating the completeness of terminal reported data based on the intersection of the actual received performance data set and the preset necessary performance data set, the optimization effect prediction unit is used for calling a pre-trained performance optimization effect prediction model, inputting the terminal actual reported data set and the full-amount terminal reported data set respectively, and outputting a predicted optimization effect and a benchmark predicted effect, and the predicted effect deviation parameter calculation unit calculates a predicted effect deviation parameter based on the predicted optimization effect and the benchmark predicted effect.

[0029] The quantitative modeling and strategy issuing module comprises a model construction unit and an optimization strategy adjustment unit, the model construction unit is used for constructing a quantitative model based on the performance data completeness parameter in a historical time period and the predicted effect deviation parameter at the corresponding moment by using a least square regression algorithm, and the optimization strategy adjustment unit is used for setting a predicted effect deviation parameter threshold, comparing the real-time predicted effect deviation with the threshold, and triggering a terminal performance adjustment action according to the comparison result.

[0030] The strategy evaluation module comprises a credibility factor calculation unit and a strategy execution unit, the credibility factor calculation unit is used for calculating a strategy credibility factor, and the strategy execution unit is used for dividing the credibility factor into intervals, counting the success rate of the strategy in each interval, determining a minimum acceptable success rate according to the business requirement, finding a minimum credibility factor meeting the success rate as an execution threshold, and executing the corresponding strategy when the credibility factor exceeds the execution threshold, otherwise, the strategy is not executed.

[0031] Compared with the prior art, the present application has the following beneficial effects:

[0032] 1. The present application calculates the completeness of terminal reported data, obtains the predicted effect by combining the pre-trained model and accounts for the deviation parameter, and then establishes a quantitative model of data completeness and predicted deviation based on historical data, which is different from the traditional method of relying on experience model and not considering the influence of data completeness on prediction, and realizes the quantitative representation of the correlation between data completeness and predicted deviation.

[0033] 2. The present application sets a predicted effect deviation threshold based on the quantitative model, and issues an optimization strategy when the deviation does not exceed the threshold, which is different from the traditional method of relying on artificial experience to set the execution condition, and provides data support through the threshold judgment mechanism.

[0034] 3. The present application collects terminal hardware resource data by dynamic patching technology, obtains application layer code path by AST analysis technology, and integrates heterogeneous data of network, business and cloud resources, which is different from the traditional method of single data source, and realizes the deep fusion of multi-source heterogeneous data. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 A flowchart of a terminal service performance monitoring optimization method based on data fusion according to the present application. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0037] In the embodiments, as shown in the drawings, the present application provides a technical solution, a terminal service performance monitoring optimization method based on data fusion, which comprises the following steps: Figure 1

[0038] Obtaining terminal hardware resource consumption data and application layer code execution path, integrating heterogeneous data of network, service and cloud resources;

[0039] According to the intersection condition of the actual received performance data set and the preset necessary performance data set, defining the performance data set completeness of the terminal reported to the cloud, using the performance data set reported by the terminal and the full-amount terminal reported data, obtaining the corresponding prediction effect through the pre-trained performance optimization effect prediction model, and calculating the prediction effect deviation parameter;

[0040] According to the performance data set completeness of the terminal reported to the cloud and the prediction effect deviation of the corresponding moment in the historical time period, establishing a quantitative model, setting a prediction effect deviation threshold, and when the deviation does not exceed the threshold, the cloud issues an optimization strategy to the terminal;

[0041] Calculating a strategy credibility factor, and executing a strategy action with reference to a strategy credibility factor threshold.

[0042] Obtaining terminal hardware resource consumption data and application layer code execution path, integrating heterogeneous data of network, service and cloud resources, and the specific steps comprise:

[0043] Collecting terminal hardware resource consumption data by using dynamic insertion technology, denoted as set H={h1, h2,..., hm}, wherein h1, h2,..., hm represent the 1st, 2nd,..., mth terminal hardware resource consumption data, and m represents the type quantity of terminal hardware resource consumption data;

[0044] ​Combining the Abstract Syntax Tree (AST) to parse the execution path of the application layer code, we denote it as the code path set C = {c1, c2, ..., cp}, where c1, c2, ..., cp represent the 1st, 2nd, ..., pth code execution paths, respectively, and p represents the total number of code execution paths;

[0045] The integration of network transmission, business process, and cloud resource type data involves the following steps: collecting network transmission data using non-intrusive bypass monitoring technology, collecting business process data using distributed transaction tracing technology, and collecting cloud resource data using virtualization layer Hypervisor monitoring probe technology.

[0046] Specifically, hardware resource consumption data of a smartphone is collected through dynamic instrumentation technology and denoted as set H = {h1 = 25% (CPU utilization), h2 = 40% (memory usage), h3 = 15% (battery power consumption)} (m = 3, 3 types of hardware data);

[0047] By combining AST technology to analyze the execution path of application layer code, we obtain the set C = {c1 = "Splash page → Homepage loading", c2 = "Homepage → Message push", c3 = "Message → Image rendering"} (p = 3, 3 core paths);

[0048] Network data (non-intrusive bypass monitoring): Network latency = 80ms, data transmission rate = 2Mbps;

[0049] Business data (distributed transaction tracing): Number of active APP users = 5000, message volume = 2000 messages / minute;

[0050] Cloud resource data (Hypervisor probe): Cloud server CPU load = 30%, storage usage = 20GB; after integration, a heterogeneous data set is formed.

[0051] Based on the intersection of the actual received performance data set and the preset necessary performance data set, the completeness of the performance data set reported by the terminal to the cloud is defined. Using the performance data set reported by the terminal and the full set of terminal-reported data, the corresponding prediction effect is obtained through a pre-trained performance optimization effect prediction model, and the prediction effect deviation parameter is calculated. The specific steps include:

[0052] The completeness of the performance data set reported by the terminal to the cloud at time ti is defined as follows: ati = (S1 / S2) * 100%;

[0053] Wherein, ati represents the completeness of the performance data set reported by the terminal to the cloud at ti, S1 represents the number of elements in the intersection of the actual cloud-received performance data set and the preset necessary performance data set; S2 represents the number of elements in the preset performance data set, and the performance data set includes terminal hardware resource consumption data, application layer code execution path, network, service and cloud resource heterogeneous data;

[0054] Based on the terminal hardware resource consumption data actually reported by the terminal to the cloud at ti, input into the pre-trained performance optimization effect prediction model, output the actual prediction optimization effect at ti, wherein the actual prediction optimization effect represents the percentage of reduction of the terminal hardware resource consumption data actually reported by the terminal to the cloud;

[0055] The terminal hardware resource consumption data reported by the terminal to the cloud at ti is input into the performance optimization effect prediction model as input, and the benchmark prediction effect at ti is output, and the prediction effect deviation parameter at ti is calculated, which is defined as follows: b = |A1-A2| / A2*100%; wherein b represents the benchmark prediction effect, A1 represents the prediction optimization effect at ti, and A2 represents the benchmark prediction effect at ti. The terminal hardware resource consumption data reported by the terminal to the cloud represents the complete terminal hardware resource consumption data in an ideal state, and the benchmark prediction effect represents the percentage of reduction of the terminal hardware resource consumption data reported by the terminal to the cloud.

[0056] Specifically, the preset necessary performance data set (which needs to contain CPU usage, memory occupation, network delay, cloud load) contains 4 types;

[0057] The actual received performance data set contains CPU usage, memory occupation, network delay, battery power consumption and message sending amount, which contains 5 types;

[0058] The intersection is {CPU usage, memory occupation, network delay}, which contains 3 types, according to the formula: ati = (3 / 4) * 100% = 75% (the data completeness at ti is 75%).

[0059] According to the completeness of the performance data set reported by the terminal to the cloud in the historical time period and the prediction effect deviation at the corresponding moment, a quantitative model is established, and the prediction effect deviation threshold is set. When the deviation does not exceed the threshold, the cloud issues optimization strategy to the terminal, and the specific steps include:

[0060] According to the completeness of the performance data set reported by the terminal to the cloud in the historical time period and the prediction effect deviation parameter at the corresponding moment, a quantitative model of data completeness to prediction effect deviation parameter is established by least square regression, which is denoted as b = f(ati) + ε; wherein b represents the prediction effect deviation parameter, f(ati) represents the fitting function, and ε represents the random error term;

[0061] A prediction effect deviation parameter threshold is set, the prediction effect deviation is compared with the set prediction effect deviation parameter threshold, and when the set prediction effect deviation parameter threshold is not exceeded, the cloud issues an optimization strategy to the terminal.

[0062] Specifically, the input terminal actually reports a data set to a pre-trained model, and an optimization effect predicted at time ti is output: CPU usage is reduced by 20% (i.e., from 25% to 20%);

[0063] The input full-quantity terminal reporting data set (containing preset necessary 4 types of data + other supplementary data) is input to the model, and a benchmark prediction effect is output: CPU usage is reduced by 30% (i.e., from 25% to 17.5%), and the deviation parameter is calculated according to the formula to be 33.3%;

[0064] The completeness a and the deviation b data of 10 historical time points are selected, a quantization model is obtained through least square regression, the prediction effect deviation threshold is set to 35%, the deviation parameter at the current time ti is 33.3% (≤ 35%), the condition is satisfied, and the cloud issues an optimization strategy to the terminal.

[0065] A strategy credibility factor is calculated, and a strategy action is performed with reference to a strategy credibility factor threshold. The specific steps include:

[0066] The strategy credibility factor is calculated, and is defined as shown below: φ = 1-b; wherein φ represents the credibility factor;

[0067] For each historical time point, the calculated credibility factor and the strategy execution result at the corresponding time point are recorded, wherein the strategy execution result includes strategy execution success and strategy execution failure, the strategy execution success is determined when the measured performance optimization effect reaches z times of the predicted performance optimization effect, and the strategy execution failure is determined when the measured performance optimization effect does not reach z times of the predicted performance optimization effect, z represents the multiple, wherein the measured performance optimization effect represents the percentage of reduction in actual observed terminal hardware resource consumption data after the terminal executes the optimization strategy issued by the cloud; and the predicted performance optimization effect represents the average value of the benchmark prediction effect at time ti and the actual prediction optimization effect at time ti.

[0068] The credibility factor is divided into intervals, the success rate of the strategy in each interval is counted, the minimum acceptable success rate is determined according to the business demand, the minimum credibility factor that meets the success rate is found as the execution threshold; when the credibility factor exceeds the execution threshold, the corresponding strategy is executed, otherwise the strategy is not executed.

[0069] Specifically, the actual optimization effect at time ti is 18%; the full-quantity data optimization effect is 25%; the deviation parameter b = 33.3%, the credibility factor is calculated by substituting the formula, and the credibility factor and the strategy execution result at the historical time point are counted.

[0070] A terminal service performance monitoring optimization system based on data fusion, comprising a data acquisition module, a performance data analysis module, a quantitative modeling and strategy issuing module, and a strategy evaluation module, the data acquisition module is used to acquire terminal hardware resource consumption data and application layer code execution path, and integrate heterogeneous data of network, service and cloud resources; the performance data analysis module is used to utilize the performance data set reported by the terminal and the full terminal reported data, obtain the corresponding prediction effect through the pre-trained performance optimization effect prediction model, and calculate the prediction effect deviation parameter; the quantitative modeling and strategy issuing module is used to establish a quantitative model according to the performance data set completeness reported by the terminal to the cloud in a historical time period and the prediction effect deviation at the corresponding moment, set a prediction effect deviation threshold, and when the deviation does not exceed the threshold, the cloud issues an optimization strategy to the terminal; the strategy evaluation module is used to calculate the strategy credibility factor according to the performance optimization effect of the actual performance data set and the full performance data set at the corresponding moment, and execute the strategy action by referring to the strategy credibility factor threshold.

[0071] The data acquisition module comprises a terminal hardware resource acquisition unit, an application layer code path analysis unit and a heterogeneous data integration unit, the terminal hardware resource acquisition unit is used to acquire terminal hardware resource consumption data by using dynamic insertion technology; the application layer code path analysis unit is used to analyze the application layer code execution path in combination with the AST abstract syntax tree; and the heterogeneous data integration unit is used to integrate network transmission, service process and cloud resource type data.

[0072] The performance data analysis module comprises a performance data completeness calculation unit, an optimization effect prediction unit and a prediction effect deviation parameter calculation unit, the performance data completeness calculation unit is used to calculate the completeness of the terminal reported data based on the intersection of the actual received performance data set and the preset necessary performance data set; the optimization effect prediction unit is used to call the pre-trained performance optimization effect prediction model, input the terminal actual reported data set and the full terminal reported data set respectively, and output the predicted optimization effect and the benchmark prediction effect; and the prediction effect deviation parameter calculation unit calculates the prediction effect deviation parameter based on the predicted optimization effect and the benchmark prediction effect.

[0073] The quantitative modeling and strategy issuing module comprises a model construction unit and an optimization strategy adjustment unit, the model construction unit is used to construct a quantitative model based on the performance data completeness parameter in a historical time period and the prediction effect deviation parameter at the corresponding moment by using the least square regression algorithm; and the optimization strategy adjustment unit is used to set a prediction effect deviation parameter threshold, compare the real-time prediction effect deviation with the threshold, and trigger the terminal performance adjustment action according to the comparison result.

[0074] The policy evaluation module comprises a credibility factor calculation unit and a policy execution unit, the credibility factor calculation unit is used for calculating a policy credibility factor according to a performance optimization effect of the actual performance data set and a performance optimization effect of the full-amount performance data set; the policy execution unit is used for dividing intervals for the credibility factor, counting a success rate of the policy in each interval, determining a minimum acceptable success rate according to a business requirement, finding a minimum credibility factor meeting the success rate as an execution threshold; when the credibility factor exceeds the execution threshold, the corresponding policy is executed, otherwise the policy is not executed.

[0075] It will be obvious to a person skilled in the art that the application is not limited to the details of the above-described exemplary embodiments but can be implemented in other embodiments without departing from the scope of the application. The embodiments are to be considered in all respects as being illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the description above, and all changes which come within the meaning and range of equivalents of the claims are therefore intended to be embraced therein. No reference signs in the claims should be considered as limiting the scope of the claims to the features to which the reference signs are attached. The word 'comprising' does not exclude the presence of elements or steps other than those listed in a claim. The word 'a' or 'an' preceding an element does not exclude the presence of a plurality of such elements. It is appreciated that features of the application that are, at present, described and claimed in alternative embodiments can be provided in combination in a single embodiment.

Claims

1. A method for monitoring and optimizing terminal service performance based on data fusion, characterized in that: The method includes the following steps: Acquire terminal hardware resource consumption data and application layer code execution paths, and integrate heterogeneous data from network, business, and cloud resources; The terminal hardware resource consumption data is collected using dynamic instrumentation technology and denoted as set H={h1,h2,...,hm}, where h1,h2,...,hm represent the 1st, 2nd,...,mth terminal hardware resource consumption data and m represents the number of types of terminal hardware resource consumption data. Combining the Abstract Syntax Tree (AST) to parse the application layer code execution path, we denote it as the code path set C={c1,c2,...,cp}, where c1,c2,...,cp represent the 1st, 2nd,...,pth code execution paths respectively, and p represents the total number of code execution paths; The specific steps for integrating network transmission, business process, and cloud resource type data are as follows: non-intrusive bypass monitoring technology is used to collect network transmission data, distributed transaction tracing technology is used to collect business process data, and virtualization layer Hypervisor monitoring probe technology is used to collect cloud resource data. Based on the intersection of the actual received performance data set and the preset necessary performance data set, the completeness of the performance data set reported by the terminal to the cloud is defined. Using the terminal hardware resource consumption data reported by the terminal to the cloud and the terminal hardware resource consumption data reported by all terminals to the cloud, the corresponding prediction effect is obtained through the pre-trained performance optimization effect prediction model, and the prediction effect deviation parameter is calculated. Among them, the terminal hardware resource consumption data reported by all terminals to the cloud represents the complete terminal hardware resource consumption data under ideal conditions, and the baseline prediction effect represents the percentage reduction of the terminal hardware resource consumption data reported by all terminals to the cloud. Based on the completeness of the performance data set reported by the terminal to the cloud within a historical time period and the deviation of the prediction effect at the corresponding time, a quantitative model is established, and a prediction effect deviation threshold is set. When the deviation does not exceed this threshold, the cloud sends optimization strategies to the terminal. Calculate the policy credibility factor and execute policy actions based on the policy credibility factor threshold.

2. The terminal service performance monitoring and optimization method based on data fusion according to claim 1, characterized in that: Based on the intersection of the actual received performance data set and the preset necessary performance data set, the completeness of the performance data set reported by the terminal to the cloud is defined. Using the performance data set reported by the terminal and the full set of terminal-reported data, the corresponding prediction effect is obtained through a pre-trained performance optimization effect prediction model, and the prediction effect deviation parameter is calculated. The specific steps include: The completeness of the performance data set reported by the terminal to the cloud at time ti is defined as follows, based on the actual received performance data set at time ti and the preset necessary performance data set: ati=(S1 / S2)*100%; Where ati represents the completeness of the performance data set reported by the terminal to the cloud at time ti, S1 represents the number of elements in the intersection of the actual performance data set received by the cloud and the preset necessary performance data set; S2 represents the number of elements in the preset performance data set, which includes heterogeneous data of terminal hardware resource consumption, application layer code execution path, network, business and cloud resources. Based on the terminal hardware resource consumption data actually reported to the cloud by the terminal at time ti, the data is input into the pre-trained performance optimization effect prediction model, and the actual predicted optimization effect at time ti is output. The actual predicted optimization effect represents the percentage reduction in the terminal hardware resource consumption data actually reported to the cloud by the terminal. The terminal hardware resource consumption data reported by all terminals to the cloud at time ti is used as input to the performance optimization effect prediction model. The output is the baseline prediction effect at time ti. The prediction effect deviation parameter at time ti is calculated and defined as follows: b=|A1-A2| / A2*100%; where b represents the baseline prediction effect, A1 represents the prediction optimization effect at time ti, A2 represents the baseline prediction effect at time ti, the terminal hardware resource consumption data reported by all terminals to the cloud represents the complete terminal hardware resource consumption data under ideal conditions, and the baseline prediction effect represents the percentage reduction of the terminal hardware resource consumption data reported by all terminals to the cloud.

3. The terminal service performance monitoring and optimization method based on data fusion according to claim 2, characterized in that: Based on the completeness of the performance data set reported by the terminal to the cloud within a historical time period and the deviation of the prediction effect at the corresponding time, a quantitative model is established, and a prediction effect deviation threshold is set. When the deviation does not exceed this threshold, the cloud sends an optimization strategy to the terminal. The specific steps include: Based on the completeness parameter of the performance data set reported by the terminal to the cloud within a historical time period and the prediction effect deviation parameter at the corresponding time, a quantitative model of the data completeness on the prediction effect deviation parameter is established by least squares regression, denoted as b=f(ati)+ε; where b represents the prediction effect deviation parameter, f(ati) represents the fitting function, and ε represents the random error term; Set a threshold for the prediction effect deviation parameter, compare the prediction effect deviation with the set threshold, and if it does not exceed the set threshold, the cloud sends an optimization strategy to the terminal.

4. The terminal service performance monitoring and optimization method based on data fusion according to claim 3, characterized in that: Calculate the policy credibility factor, and execute policy actions based on the policy credibility factor threshold. The specific steps include: The confidence factor of the strategy is calculated and defined as follows: φ = 1 - b; where φ represents the confidence factor. For each historical moment, the confidence factor and strategy execution result calculated at that moment are recorded. The strategy execution result includes strategy execution success and strategy execution failure. At time ti, if the measured performance optimization effect reaches z times the predicted performance optimization effect, it is determined that the strategy execution is successful. If the measured performance optimization effect does not reach z times the predicted performance optimization effect, it is determined that the strategy execution fails. z represents the multiple. The measured performance optimization effect represents the percentage reduction in the actual observed terminal hardware resource consumption data after the terminal executes the optimization strategy issued by the cloud. The predicted performance optimization effect represents the average of the baseline prediction effect at time ti and the actual predicted optimization effect at time ti. The credibility factor is divided into intervals, and the success rate of the strategy within each interval is calculated. The minimum acceptable success rate is determined based on business needs, and the minimum credibility factor that meets this success rate is found as the execution threshold. If the credibility factor exceeds the execution threshold, the corresponding strategy is executed; otherwise, the strategy is not executed.

5. A terminal service performance monitoring and optimization system based on data fusion, applied to the terminal service performance monitoring and optimization method based on data fusion as described in any one of claims 1-4, characterized in that: The system includes: a data acquisition module, a performance data analysis module, a quantitative modeling and strategy delivery module, and a strategy evaluation module. The data acquisition module acquires terminal hardware resource consumption data and application layer code execution paths, integrating heterogeneous data from network, business, and cloud resources. The performance data analysis module uses the performance data set reported by the terminal and all terminal-reported data to obtain corresponding prediction results through a pre-trained performance optimization effect prediction model, and calculates prediction effect deviation parameters. The quantitative modeling and strategy delivery module establishes a quantitative model based on the completeness of the performance data set reported by the terminal to the cloud within a historical time period and the prediction effect deviation at the corresponding time, sets a prediction effect deviation threshold, and when the deviation does not exceed this threshold, the cloud delivers optimization strategies to the terminal. The strategy evaluation module calculates a strategy credibility factor based on the performance optimization effects of the actual performance data set and the full performance data set at the corresponding time, and executes strategy actions with reference to the strategy credibility factor threshold.

6. The terminal service performance monitoring and optimization system based on data fusion according to claim 5, characterized in that: The data acquisition module includes a terminal hardware resource acquisition unit, an application layer code path parsing unit, and a heterogeneous data integration unit. The terminal hardware resource acquisition unit is used to collect terminal hardware resource consumption data using dynamic instrumentation technology. The application layer code path parsing unit is used to parse the application layer code execution path in conjunction with the AST abstract syntax tree; the heterogeneous data integration unit is used to integrate network transmission, business process and cloud resource type data.

7. The terminal service performance monitoring and optimization system based on data fusion according to claim 6, characterized in that: The performance data analysis module includes a performance data integrity calculation unit, an optimization effect prediction unit, and a prediction effect deviation parameter calculation unit. The performance data integrity calculation unit calculates the integrity of the data reported by the terminal based on the intersection of the actual received performance data set and the preset necessary performance data set. The optimization effect prediction unit calls a pre-trained performance optimization effect prediction model, inputs the actual data set reported by the terminal and the full set of terminal-reported data, and outputs the predicted optimization effect and the baseline prediction effect. The prediction effect deviation parameter calculation unit calculates the prediction effect deviation parameter based on the predicted optimization effect and the baseline prediction effect.

8. The terminal service performance monitoring and optimization system based on data fusion according to claim 7, characterized in that: The quantitative modeling and strategy delivery module includes a model building unit and an optimization strategy adjustment unit. The model building unit is used to construct a quantitative model based on the performance data completeness parameter within a historical time period and the prediction effect deviation parameter at the corresponding time using the least squares regression algorithm. The optimization strategy adjustment unit is used to set a threshold for the prediction effect deviation parameter, compare the real-time prediction effect deviation with the threshold, and trigger terminal performance adjustment actions based on the comparison results.

9. The terminal service performance monitoring and optimization system based on data fusion according to claim 8, characterized in that: The strategy evaluation module includes a credibility factor calculation unit and a strategy execution unit. The credibility factor calculation unit is used to calculate the strategy credibility factor. The strategy execution unit is used to divide the credibility factor into intervals, calculate the success rate of the strategy in each interval, determine the minimum acceptable success rate according to business needs, and find the minimum credibility factor that meets the success rate as the execution threshold; if the credibility factor exceeds the execution threshold, the corresponding strategy is executed, otherwise the strategy is not executed.

Images