A software integration level monitoring system based on a modular architecture

Abstract

The application discloses a software integration degree monitoring system based on a modular architecture, and particularly relates to the field of software integration degree situation awareness, and comprises a data input module, a task distribution module, a path tracking module, a voting module and an evaluation module; the data input module inputs data contained in software as input variables through a data access pipeline, converts the data into standardized input blocks, simultaneously establishes a redundant data mapping, and makes the same data reused in multiple calculation paths; through the redundant data mapping and a task hash mechanism, the same data is reused in multiple calculation paths, and hash consistency verification is performed when the calculation paths are executed; compared with a traditional external checking mode, the present scheme utilizes redundant information generated in a calculation process to perform real-time error detection, thereby effectively reducing calculation deviation and data pollution risk in a high-concurrency calculation and distributed storage environment, and improving the stability of the overall system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of software integration situation awareness, and more particularly, to a software integration monitoring system based on a modular architecture. BACKGROUND

[0002] In the current technology, software integration refers to the closeness of data sharing and functional coupling between modules, that is, an index for measuring the internal coordination efficiency of the system; and situation awareness is the real-time monitoring and identification ability of the system to its own and external environment state, which is used to capture abnormal information in time; software integration situation awareness refers to real-time monitoring of the data interaction and collaborative integration state between software modules in the system, discovering potential abnormalities and integration risks in time by dynamically evaluating interface stability and data flow smoothness, so as to provide a basis for early warning;

[0003] In high-concurrency computing, distributed storage or real-time data processing, the system usually relies on software integration to ensure the smoothness and accuracy of its data operation. Since the dependence of software on data consistency and error detection mainly concentrates on external checking mechanisms rather than directly utilizing the redundancy information in software operation, some errors cannot be captured in real time, especially in high-throughput systems. The lack of operation redundancy as an error detection means for software integration will cause data pollution, calculation deviation or system crash, affecting the overall reliability. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a software integration monitoring system based on a modular architecture, which maps redundant data and uses a task hashing mechanism to reuse the same data in multiple computing paths and perform hash consistency verification when the computing path is executed, so as to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a software integration monitoring system based on a modular architecture, comprising a data input module, a task distribution module, a path tracking module, a voting module and an evaluation module.

[0006] The data input module inputs the data contained in the software as input variables through the establishment of a data access pipeline, converts them into standardized input blocks, and establishes a redundant data mapping to reuse the same data in multiple computing paths;

[0007] The data input module generates a unique task hash for each input block, constructs an input parsing stack, and matches a calculation algorithm according to the data type and distributes it to a heterogeneous computing pool;

[0008] The task distribution module distributes the standardized input blocks to multiple different algorithm models for performing calculations by setting up a heterogeneous computing distribution index table; a multi-layer calculation buffer is established at each algorithm model calculation node to store original input data, intermediate results and redundant calculation data in the calculation process, respectively;

[0009] The path tracking module is used to record the calculation execution process and establish a calculation path tracing model based thereon;

[0010] The voting module is used to store the calculation outputs of each algorithm model and establish a voting mechanism based on the comparison of the redundant calculation results, so as to optimize the reliability of the calculation results through the voting mechanism;

[0011] The evaluation module quantifies the integration level of the software based on the calculation history of the algorithm model, the reliability calculated by the voting mechanism and the situation awareness stability of the calculation path of the calculation path tracing model.

[0012] In a preferred embodiment, the data input module, when executed, sets up an input buffer for a data access pipeline, encapsulates the original data in the software system as an input data block and appends a timestamp index, parses the input data block through a standardized input conversion layer, and converts it into a standardized input block according to the data format and category;

[0013] The redundant data mapping construction layer of the data input module introduces a redundant data identification table in the standardized input block creation process and tracks the reuse relationship of the same data in multiple calculation paths through hash mapping;

[0014] The input parsing stack of the data input module parses the type, characteristics and calculation requirements of the input data and constructs an input parsing stack index table, which layers the input blocks according to data attributes; the redundant calculation path distribution distributes multiple copies of the same standardized input block according to different algorithm paths based on the index information of the input parsing stack, so that the same data is redundantly calculated in multiple calculation processes;

[0015] The data input module sets up a calculation path index table to record the distribution trajectory of the standardized input block in different calculation nodes, forming a traceable calculation path mapping; based on the calculation path index table, tasks are dynamically allocated in the heterogeneous computing pool, data consistency verification is performed through the data input module, a task index table is constructed before the calculation task starts to compare the task hash of the current input block, so that the redundant data is calculated consistently in each path.

[0016] In a preferred embodiment, when the computing task enters the path tracking unit, a computing path index table is created to store the task identifier, computing branch information and timestamp in a hash index manner; a computing path tracing model is constructed by the path tracking unit, which constructs three execution vectors as inputs based on the task execution trajectory, computing result deviation and algorithm stability to evaluate the integrity and abnormality of the computing path;

[0017] When the computing task in the path tracking unit is executed, a computing path snapshot is generated, which includes the input hash, computing state and computing output hash of the current computing step, and is stored in the path tracing index table.

[0018] In a preferred embodiment, the task execution trajectory included in the path tracking unit is recorded in the form of a computing step chain table, each computing step being associated with input data hash, execution algorithm identifier and output data hash; the computing result deviation is quantified according to a computing error matrix, the "rows" of the computing error matrix representing the results of the same task on different computing paths, and the "columns" of the computing error matrix representing the computing deviations of the task at different computing time points, based on which the computing error evolution trend across the time dimension is constructed;

[0019] The algorithm stability included in the path tracking unit is evaluated by the repeatability and error convergence of the historical computing path; if the results presented in different execution batches of the same computing task are highly consistent, the stability score of the corresponding algorithm is determined to be increased, otherwise it is marked as an abnormal computing path.

[0020] In a preferred embodiment, the stability score of the computing path is output based on the result of the computing path tracing model, and if the decay trend acceleration rate of the stability score of a computing path is greater than the decay trend acceleration rate threshold preset by the system, the trustworthiness of the computing path in the system is determined to be abnormal;

[0021] When the trustworthiness of the computing path is determined to be abnormal, the redundant computing path is activated to trigger the standby algorithm to execute the same task and perform comparison with the original computing path result; the analysis result of the computing path tracing model is finally fed back to the evaluation module, and the software integration level is analyzed according to the situation awareness stability of the computing path.

[0022] In a preferred embodiment, after the completion of the computing task, the computing outputs of each algorithm model are stored in the voting module and indexed and stored according to the task identifier; when storing the computing result, the corresponding redundant computing data is also stored, including the intermediate variables, error range and execution time calculated by different algorithms under the same input condition, to construct a redundant computing comparison benchmark;

[0023] The input data structure of the voting mechanism of the voting module includes a calculation result, a redundant calculation comparison value and a historical calculation path record group; a voting matrix is constructed based on the voting module, all calculation outputs are compared for redundancy, the consistency of redundant data in different calculation paths is quantified based on a calculation path index table, and a multi-dimensional voting decision matrix is output.

[0024] In a preferred embodiment, a voting decision layer is constructed based on the voting mechanism, the voting decision layer performs weighted voting on the calculation results, different weights are given to different calculation results according to historical calculation stability factors, redundant data consistency factors and calculation path integration score factors, an abnormality detection mechanism is established, when the calculation result of a certain algorithm model deviates from the redundant comparison data by more than a preset deviation threshold, the voting weight of the calculation path is reduced, and the abnormal information is stored in the calculation path index table.

[0025] Based on the calculation result consistency, the redundant data deviation and the historical calculation trajectory, the reliability score of the final voting decision is evaluated and output.

[0026] Technical effects and advantages of the present application:

[0027] 1. Through the redundant data mapping and task hash mechanism, the same data is reused in multiple calculation paths, and hash consistency verification is performed when the calculation path is executed. Compared with the traditional external checking method, the present scheme uses the redundant information generated in the calculation process for real-time error detection, thereby effectively reducing the calculation deviation and data pollution risk in the high-concurrency calculation and distributed storage environment, and improving the stability of the overall system.

[0028] 2. By introducing the path tracking module, a calculation path tracing model is established, the execution trajectory, error change trend and algorithm stability of the calculation path are dynamically analyzed, and the calculation path snapshot and index table are used to make the execution process of all calculation tasks traceable.

[0029] 3. By storing the output results of each calculation path and using redundant calculation data comparison, a multi-dimensional voting decision matrix is constructed, the voting mechanism integrates the calculation history, the calculation path score and the redundant comparison result to give weights to the calculation results of different paths, realizes dynamic reliability evaluation, and eliminates abnormal calculation results.

[0030] 4. By introducing calculation history analysis in the evaluation module, the voting mechanism score and the calculation path tracing score are integrated to facilitate long-term trend analysis of the calculation path.

[0031] 5. By combining path tracing, calculation history analysis and voting reliability score in the evaluation module, a situation awareness stability model of the calculation path is constructed, the software integration level is quantified, and the calculation management in the high-concurrency calculation and distributed storage environment is relatively more efficient. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 System module diagram for the present invention.

[0033] Figure 2 System execution flow for the present invention Figure One .

[0034] Figure 3 System execution flow for the present invention Figure Two .

[0035] Figure 4 System execution flow for the present invention Figure Three .

[0036] Figure 5 System execution flow for the present invention Figure Four .

[0037] Figure 6 Overall flow architecture diagram for the present invention. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only 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.

[0039] With reference to the drawings in the description Figures 1-6 , an embodiment of the present application is a software integration monitoring system based on a modular architecture, which comprises a data input module, a task distribution module, a path tracking module, a voting module and an evaluation module.

[0040] The data input module inputs the data contained in the software as input variables through the established data access pipeline, converts the data into standardized input blocks, and establishes a redundant data mapping to reuse the same data in multiple calculation paths.

[0041] The data input module generates a unique task hash for each input block, which is used to track the data flow and ensure calculation consistency. An input parsing stack is constructed, which matches appropriate calculation algorithms according to the data type and distributes them to a heterogeneous computing pool.

[0042] The task distribution module distributes the standardized input blocks to multiple different algorithm models for calculation by establishing a heterogeneous computing distribution index table. A multi-layer calculation buffer is established at each algorithm model calculation node to store the original input data, intermediate results and redundant calculation data in the calculation process.

[0043] The path tracking module is used to record the execution process of the calculation and establish a calculation path tracing model based on the same;

[0044] The voting module is used to store the calculation output of each algorithm model and establish a voting mechanism based on the comparison of the redundant calculation results, so as to optimize the reliability of the calculation results through the voting mechanism;

[0045] The evaluation module quantifies the integration level of the software based on the calculation history of the algorithm model, the reliability calculated by the voting mechanism, and the situation awareness stability of the calculation path of the calculation path tracing model.

[0046] The data input module, when executed, sets up an input buffer for the data access pipeline, encapsulates the original data in the software system as an input data block, and appends a timestamp index. The input data block is parsed through a standardized input conversion layer, and is converted into a standardized input block according to the data format and category;

[0047] The redundant data mapping construction layer of the data input module introduces a redundant data identification table during the creation of the standardized input block, and tracks the reuse relationship of the same data in multiple calculation paths through hash mapping;

[0048] The input parsing stack of the data input module parses the type, characteristics, and calculation requirements of the input data, and constructs an input parsing stack index table. The input parsing stack index table layers the input block according to data attributes. The redundant calculation path distribution distributes multiple copies of the same standardized input block according to different algorithm paths based on the index information of the input parsing stack, so that the same data is redundantly calculated in multiple calculation processes;

[0049] The data input module sets up a calculation path index table to record the distribution trajectory of the standardized input block in different calculation nodes, forming a traceable calculation path mapping. According to the calculation path index table, tasks are dynamically allocated within the heterogeneous computing pool, and data consistency verification is performed through the data input module. Before the calculation task starts, a task index table is constructed to compare the task hash of the current input block, so that the redundant data is calculated consistently in each path.

[0050] When the calculation task enters the path tracking unit, a calculation path index table is created to store the task identification, calculation branch information, and timestamp in a hash index manner. The calculation path tracing model is constructed through the path tracking unit. The calculation path tracing model constructs three execution vectors as inputs based on the task execution trajectory, calculation result deviation, and algorithm stability to evaluate the integrity and abnormality of the calculation path;

[0051] When the calculation task in the path tracking unit is executed, a calculation path snapshot is generated, which includes the input hash, calculation state, and calculation output hash of the current calculation step, and is stored in the path tracing index table, which is used for subsequent anomaly detection.

[0052] The task execution track included in the path tracking unit is recorded in the form of a chain table of calculation steps, each calculation step being associated with an input data hash, an execution algorithm identifier, and an output data hash, the purpose being to make the path completely traceable; the calculation result deviation is quantified according to a calculation error matrix, the “rows” of the calculation error matrix representing the results of the same task on different calculation paths, and the “columns” of the calculation error matrix representing the calculation deviations of the task at different calculation time points, based on which an evolution trend of the calculation error across the time dimension is constructed;

[0053] The algorithm stability included in the path tracking unit is evaluated by the repeatability and error convergence of historical calculation paths; if the results presented in different execution batches of the same calculation task are highly consistent, it is determined that the stability score of the corresponding algorithm is increased, otherwise it is marked as an abnormal calculation path.

[0054] The stability score of the calculation path is output based on the results of the calculation path tracing model, and if the acceleration rate of the decay trend of the stability score of a certain calculation path is greater than the decay trend acceleration rate threshold preset by the system, it is determined that the trustworthiness of the calculation path in the system is abnormal, that is, the calculation result may be affected by abnormal factors;

[0055] When it is determined that the trustworthiness of the calculation path is abnormal, the redundant calculation path is activated, the standby algorithm is triggered to execute the same task, and the results are compared with the original calculation path; the analysis results of the calculation path tracing model are finally fed back to the evaluation module, the software integration level is analyzed according to the situation awareness stability of the calculation path, and the scheduling strategy of the calculation task is adjusted.

[0056] After the completion of the calculation task, the calculation outputs of each algorithm model are stored in the voting module and are stored according to the task identifier; when storing the calculation results, the corresponding redundant calculation data is also stored, including intermediate variables, error ranges, and execution times calculated by different algorithms under the same input conditions, to construct a redundant calculation comparison benchmark;

[0057] The input data structure of the voting mechanism of the voting module includes calculation results, redundant calculation comparison values, and historical calculation path records; a voting matrix is constructed based on the voting module, redundant comparisons are performed on all calculation outputs, the consistency of redundant data in different calculation paths is quantified based on the calculation path index table, and a multi-dimensional voting decision matrix is output.

[0058] A voting decision layer is constructed based on a voting mechanism. The voting decision layer performs weighted voting on the calculation results, assigning different weights to different calculation results based on historical calculation stability factors, redundant data consistency factors, and calculation path integration degree scoring factors. An anomaly detection mechanism is established. When the deviation between the calculation result of a certain algorithm model and the redundant comparison data exceeds a preset deviation threshold, the voting weight of that calculation path is reduced, and the anomaly information is stored in the calculation path index table.

[0059] Based on the consistency of calculation results, the deviation of redundant data, and the historical calculation trajectory, the credibility score of the final voting decision is evaluated and output.

[0060] As a further step, it should be noted that:

[0061] A computational path tracing model is built based on the path tracing module. The computational path tracing model is used to analyze the execution of the computational path, including the stability of the computational snapshot, path deviation, time decay factors, and constructs a global stability index for the computational path.

[0062] In calculating the snapshot stability score, P i Let be used to represent the snapshot stability score of algorithm i in the computation path tracing, reflecting the local consistency of its computation path;

[0063]

[0064] Where M i S represents the total number of snapshots in the path snapshot calculation by algorithm i, which is also the number of samples in the path record; i,j S scores the computation path of algorithm i at snapshot index j. i,j This snapshot is used to reflect the stability of the computation process, such as the degree of matching between the input hash and the output hash. Let i be the median of all snapshot scores for algorithm i, in practical applications It can be used as a benchmark value for stability; λ P To calculate the deviation impact factor of the path snapshot, Used to control the degree of fluctuation in scores; W(S i,j ) represents the weighting function used to adjust the importance of different snapshots; W(·) is the Lambert-W function, used in practical applications to describe non-linear growth; δ P For weighting adjustment parameters, more specifically, δ P It is used to control the impact of snapshot scores in the calculation path;

[0065] The overall score of the computational path tracing model is quantified based on the stability score of the computational snapshot.

[0066]

[0067] where E P,i is the stability score of individual algorithm i in the computing path tracing model; κ P is the path score nonlinear amplification factor, κ P is used to affect the convergence rate of computing path fluctuation; η P is the path score exponential adjustment parameter, which is used to affect the degree of influence of snapshot score on global path stability score; E P is the global computing path tracing model score, that is, finally reflects the stability of the entire computing path in all algorithm models; N is the total number of algorithm models participating in computing;

[0068] The computing history of each algorithm model is described; the abnormal accumulation, computing complexity and consistency of each algorithm model in the historical task are analyzed to evaluate the influence of long-term computing behavior on the system;

[0069] The computing history abnormal integral is calculated, and H i is the computing history abnormal integral of algorithm i in the time interval [T0, T]; the computing data complexity correction is calculated, and D i is the computing data complexity correction coefficient of algorithm i;

[0070]

[0071] where is the instantaneous error value calculated by algorithm i at time t, which is derived from computing path deviation, redundant data comparison result, etc.; δ H is the abnormal data sensitivity factor, which is used to control the degree of influence of abnormal integral on historical computing; X i is the complexity measure of input data; K i is the task consistency score, which is obtained by historical task hash matching; 1F1(μ;ν;X i ) is the conjugate hypergeometric function, which is used to measure the influence of input data on computing history abnormal integral;

[0072] The computing history score is calculated based on H i and D i ;

[0073]

[0074] where E H,i is the computing history score of algorithm i; Γ(x) is the gamma function, which is used to describe the historical accumulation effect of nonlinear amplification of abnormal data; α H ,β H are used to control the weight of computing history data.

[0075] The global computing path tracing model score EH,i The description history global score is described;

[0076]

[0077] wherein E H is the global calculation history score; N is the total number of algorithm models participating in the calculation; is the product of the calculation history scores of all algorithms, ensuring the nonlinear amplification of abnormal scores; is the geometric mean operation index, used to balance the numerical value of the comprehensive score, and not to affect the overall result due to an abnormal score of a certain algorithm;

[0078] The reliability calculated by the voting mechanism is calculated by comprehensively calculating the redundancy calculation results, time fluctuations, and historical stability, and provides a reliability score of the calculation task;

[0079]

[0080] wherein E V,i is the voting reliability score calculated by algorithm i, E V,i is used to measure the reliability of the algorithm after redundancy calculation comparison; exp(·) is an exponential function, used to ensure that the reliability score is always positive and to enhance the influence of high reliability votes; γ V is the amplification factor of the reliability score, used to control the range of the reliability score, making it easier to distinguish low reliability scores and more significant high reliability scores in actual applications; Ei(·) is an exponential integral function, used to enhance the response of the reliability score to extreme high values, making the reliability score relatively stable; V i is the original voting score of algorithm i, derived from the matching degree of the calculation result and the redundancy comparison; δ V is the exponential adjustment parameter of the voting reliability score, δ V is used to control the sensitivity of the reliability score to the original voting score;

[0081] In addition, the 2F1(ρ,σ;τ;R i ) part in the above formula is the Gauss hypergeometric function, which is used to measure the nonlinear contribution of redundancy calculation data to the voting reliability; ρ is the hypergeometric parameter 1, controlling the main influence term of the redundancy calculation data on the reliability score; σ is the hypergeometric parameter 2, adjusting the rate of change of the reliability score under different redundancy calculation conditions; τ is the hypergeometric parameter 3, used to affect the weight of the calculation reliability on different calculation paths; R i is the redundancy calculation data index calculated by algorithm i, which includes the weighted sum of redundancy intermediate variables, error ranges, execution times, and other factors generated by the algorithm during the calculation; T i is the execution time fluctuation factor of algorithm i, T iUsed to reflect the computation time instability of the algorithm in historical tasks; lnT i This indicates taking the logarithm of the execution time to ensure it still has a stable impact on the credibility score even under extreme computation time variations; 1+|lnT i The addition of 1 in | is to prevent the logarithmic function from taking a value close to 0, ensuring that the formula can be calculated under all execution time conditions;

[0082] Ultimately based on E V,i Calculate the global voting credibility score E V E V It describes the non-linear aggregation result of the vote credibility scores calculated by all algorithms;

[0083]

[0084] in This represents the summation of the voting credibility scores for all algorithm models; ξ V The power mean exponent parameter is used to control the fusion method of voting credibility scores; To normalize the power mean, the nonlinear contributions of different voting scores are kept reasonable;

[0085] In the description of the situational awareness stability of the computation path in the computation path tracing model; through I 最终 The situational awareness stability score of the computation path is used to describe the final output of the computation path tracing model, historical computation, and voting credibility score, which measures the stability of the overall software integration of the system.

[0086]

[0087] Where exp(·) is an exponential function used to ensure that the final score is always positive and to effectively smooth the calculated credibility score; Φ H To calculate the historical transformation index, which measures the stability performance of each algorithm model in historical calculations; E H Calculate historical scores globally; α I A nonlinear adjustment parameter is used to calculate the historical scores, controlling the contribution of the historical scores to the overall stability; Φ V β is the converted vote credibility value, used to measure the impact of redundant calculation results on credibility; Li3(·) is a third-order logarithmic function, used in the above formula to enhance the nonlinear characteristics of credibility scoring, enabling it to stably provide reliable scores under different calculation conditions; I Φ is the weighting parameter for the credibility score; P To calculate the stability factor for path tracing;

[0088] Additionally, regarding ΦP arcsin(·) in the formula represents the inverse sine function, which is used to keep the calculation path stability score within a reasonable range, avoiding extreme calculation path scores affecting the overall software integration evaluation; γ I is a weight parameter for the calculation path score, which is used to control the contribution of the calculation path score to the overall software integration; in addition, 1 + Φ P part of the formula is used to ensure that the contribution of the calculation path score is within a controllable range, avoiding distortion of the final score caused by excessively large or small calculation path scores; is an overall adjustment parameter, used to control the output scale of the final score;

[0089] In addition, it needs to be explained for this system that the purpose of this scheme is to build a software integration monitoring system, so that the calculation task remains consistent and traceable on different calculation paths; it realizes accurate tracking, error detection and optimization of the calculation process through data input, task distribution, path tracking, voting decision and evaluation feedback;

[0090] In the data input stage, the system receives input data through the data access pipeline and converts it into standardized input blocks, so as to ensure that all calculation tasks are processed based on unified format data; at the same time, the scheme introduces a redundant data mapping mechanism, so that the same data can be reused by multiple calculation paths, so as to cross-verify in subsequent calculation and improve the consistency of calculation results;

[0091] In the task distribution stage, each standardized input block is assigned a unique task hash value, which is not only used for data tracking, but also makes the calculation consistency of the same data in different calculation paths; at the same time, the system distributes tasks to heterogeneous computing pools according to data types and calculation requirements, and manages calculation paths through index table to realize efficient scheduling of tasks;

[0092] After entering the path tracking stage, the system records the calculation execution process in real time and stores the calculation path for subsequent error analysis and anomaly detection; the calculation path tracing model relies on task execution trajectory, calculation result deviation and algorithm stability to evaluate the stability and credibility of the calculation path; in addition, the system also includes generating calculation snapshots, which are used to store input hash, calculation state and calculation output hash to ensure traceability of the calculation process;

[0093] The voting module is used to optimize the credibility of the calculation result, stores the calculation output of each calculation path, and establishes a voting mechanism based on the comparison of redundant calculation results; the redundant data calculated by different algorithm paths will be compared, and the final calculation result will be optimized through the voting mechanism to ensure that the overall calculation result still has credibility even if some paths are disturbed in the calculation process;

[0094] The evaluation module is the decision core part of the system, which quantifies the software integration level according to the algorithm calculation history, the reliability calculated by the voting mechanism and the calculation path tracing model, not only evaluates the quality of the current calculation task, but also optimizes the scheduling strategy of the subsequent calculation task based on the historical calculation behavior and abnormal analysis, for example, when the score of a certain calculation path continues to decline, the system can trigger the standby algorithm to calculate to improve the stability of the calculation.

[0095] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A software integration monitoring system based on a modular architecture, comprising a data input module, a task distribution module, a path tracking module, a voting module, and an evaluation module; Its features are: The data input module takes the data contained in the software as input variables by setting up a data access pipeline, converts it into standardized input blocks, and establishes redundant data mapping so that the same data can be reused in multiple computing paths. The data input module generates a unique task hash for each input block, constructs an input parsing stack, matches the calculation algorithm according to the data type, and distributes it to the heterogeneous computing pool. The task distribution module establishes a heterogeneous computing distribution index table to distribute standardized input blocks to multiple different algorithm models for computation; and establishes multi-layer computing buffers at each algorithm model computing node to store the original input data and intermediate results and redundant computing data during the computation process. The path tracing module is used to record the computation execution process and build a computation path tracing model based on it; The voting module is used to store the computational output of each algorithm model and establish a voting mechanism based on the comparison of its redundant computational results, thereby optimizing the credibility of the computational results through the voting mechanism. The evaluation module assesses the computational history of the algorithm model, the credibility of the calculations based on the voting mechanism, the stability of the computational path tracing model's situational awareness of the computational path, and the integration level of the software.

2. The software integration monitoring system based on a modular architecture according to claim 1, characterized in that: When the data input module is executed, it sets up an input buffer for the data access pipeline, encapsulates the raw data in the software system into input data blocks, adds a timestamp index, and parses the input data blocks through the standardized input conversion layer, converting them into standardized input blocks according to the data format and category; In the process of creating standardized input blocks, the redundant data mapping construction layer of the data input module introduces a redundant data identification table and tracks the reuse relationship of the same data in multiple computing paths through hash mapping. The input parsing stack of the data input module parses the type, characteristics, and computational requirements of the input data, and constructs an input parsing stack index table. The input parsing stack index table hierarchically divides the input blocks according to data attributes. The redundant computation path distribution is based on the index information of the input parsing stack. Multiple copies of the same standardized input block are distributed according to different algorithm paths, so that the same data is redundantly computed in multiple computation processes. A computation path index table is established through the data input module to record the distribution trajectory of standardized input blocks on different computing nodes, forming a traceable computation path mapping. Based on the computation path index table, tasks are dynamically allocated in the heterogeneous computing pool. Data consistency verification is performed through the data input module. Before the computation task starts, a task index table is built to compare the task hash of the current input block, so that the computation of redundant data is consistent in each path.

3. The software integration monitoring system based on a modular architecture according to claim 2, characterized in that: When a computation task enters the path tracing unit, a computation path index table is created to store the task identifier, computation branch information, and timestamp using a hash index. A computation path tracing model is constructed through the path tracing unit. The computation path tracing model uses three execution vectors as inputs based on the task execution trajectory, computation result deviation, and algorithm stability to evaluate the integrity and anomalies of the computation path. When a computation task is executed in the path tracing unit, a computation path snapshot is generated. The computation path snapshot includes the input hash, computation status, and computation output hash of the current computation step, and is stored in the path tracing index table.

4. The software integration monitoring system based on a modular architecture according to claim 3, characterized in that: The task execution trajectory included in the path tracking unit is recorded in the form of a linked list of calculation steps. Each calculation step is associated with the input data hash, the execution algorithm identifier, and the output data hash. The deviation of the calculation results is quantified based on the calculation error matrix. The "rows" of the calculation error matrix represent the results of the same task on different calculation paths, and the "columns" of the calculation error matrix represent the calculation deviation of the task at different calculation time points. Based on this, the evolution trend of calculation error across time dimensions is constructed. The stability of the algorithm included in the path tracking unit is evaluated by the repeatability and error convergence of historical computation paths. If the results presented in different execution batches of the same computation task are highly consistent, the stability score of the corresponding algorithm is increased; otherwise, it is marked as an abnormal computation path.

5. The software integration monitoring system based on a modular architecture according to claim 4, characterized in that: Based on the results of the computation path tracing model, the stability score of the computation path is output. If the rate of acceleration of the decay trend of the stability score of a certain computation path is greater than the preset threshold of the rate of acceleration of the decay trend, the trust level of the computation path in the system is determined to be abnormal. When the trust level of the computation path is determined to be abnormal, the redundant computation path is activated, the backup algorithm is triggered to perform the same task, and the result is compared with that of the original computation path. The analysis results of the computation path tracing model are ultimately fed back to the evaluation module, which analyzes the software integration level based on the situational awareness stability of the computation path.

6. A software integration monitoring system based on a modular architecture according to claim 5, characterized in that: After the computation task is completed, the computation output of each algorithm model is stored in the voting module and indexed according to the task identifier. When storing the computation results, the corresponding redundant computation data is also stored. The redundant computation data includes intermediate variables, error ranges and execution times calculated by different algorithms under the same input conditions, so as to build a redundant computation comparison benchmark. The input data structure of the voting mechanism in the voting module consists of calculation results, redundant calculation comparison values, and historical calculation path records. A voting matrix is ​​constructed based on the voting module. Redundancy comparison is performed on all calculation outputs. Based on the calculation path index table, the consistency of redundant data in different calculation paths is quantified, and a multi-dimensional voting decision matrix is ​​output.

7. A software integration monitoring system based on a modular architecture according to claim 6, characterized in that: A voting decision-making layer is constructed based on a voting mechanism. The voting decision-making layer performs weighted voting on the calculation results, assigning different weights to different calculation results based on factors such as historical calculation stability, redundancy data consistency, and calculation path integration score. An anomaly detection mechanism is established. When the deviation between the calculation result of a certain algorithm model and the redundant comparison data exceeds a preset deviation threshold, the voting weight of that calculation path is reduced, and the anomaly information is stored in the calculation path index table. Based on the consistency of calculation results, the deviation of redundant data, and the historical calculation trajectory, the credibility score of the final voting decision is evaluated and output.

Images