A Big Data-Based Enterprise Information Technology Service Consulting System
By utilizing a big data-based enterprise IT service consulting system, differential storage and adaptive evidence collection technologies were employed to address the issues of inconsistent data status and missing evidence in the enterprise IT environment, thereby achieving efficient verification of consulting results and improved reliability.
Patent Information
- Application Number
- CN202611319269.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-28
- Publication Date
- 2026-11-17
AI Technical Summary
Existing online consulting systems struggle to establish verifiable correspondences between real-time topology, operational load, and fault status within enterprise IT environments. This leads to increased network transmission overhead, storage consumption, and inference latency as data volume grows. Furthermore, the lack of critical operational evidence and the absence of target environment verification in consulting solutions reduce reliability.
Employing a multi-source data acquisition gateway, a differential state graph engine, a consultation task subgraph generator, an adaptive evidence acquisition scheduler, a candidate solution generator, a sandbox verification engine, and a resource orchestrator, this system leverages differential storage and adaptive evidence acquisition to verify the impact of changes on operational metrics using a local digital twin sandbox, thereby improving the traceability and verifiability of consultation results.
It reduces the amount of data transmission and computation involved in full synchronization and repeated calculations of multi-source data, controls disturbances in the production system, improves the reliability and traceability of consulting results, and reduces the risk of trial and error in the production environment.
Smart Images

Figure CN123115474A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of big data analytics, and in particular to a big data-based enterprise information technology service consulting system. Background Technology
[0002] An enterprise IT environment typically consists of servers, network devices, databases, middleware, container platforms, business applications, cloud resources, and security devices. Configuration data, performance metrics, logs, work orders, call chains, and change records for different components are scattered across multiple systems, exhibiting characteristics such as heterogeneous data structures, varying update frequencies, different time validity periods, and potential conflicts.
[0003] Existing online consultation or recommendation systems often return consultation content based on keywords, user profiles, or static knowledge bases, making it difficult to establish a verifiable correspondence between consultation conclusions and the real-time topology, operational load, and fault status of the enterprise's IT environment. As data volume increases, full synchronization and full recalculation also lead to increased network transmission overhead, storage usage, and inference latency.
[0004] Furthermore, companies often lack crucial operational evidence during the consulting process. Existing systems typically require the collection of large amounts of data at once, or provide solutions based on incomplete data. The former disrupts the production system and increases data transmission volume, while the latter reduces the reliability of the solution. Solutions generated solely based on similar historical cases also lack operational validation in the target company's environment, potentially violating service level objectives or introducing new resource bottlenecks.
[0005] Therefore, there is a need for a technical solution for enterprise information technology components and operational data that can reduce redundant transmission and repetitive calculations in the context of continuously changing multi-source heterogeneous data, selectively supplement evidence according to its contribution to the uncertainty of the results, and verify the impact of technical changes on operational indicators through an isolated sandbox, thereby improving the traceability, verifiability and predictive reliability of consulting results. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention provides a big data-based enterprise information technology service consulting system to solve problems such as inconsistent data status from multiple sources, high consumption of full-scale computing resources, lack of key evidence, and lack of target environment verification for consulting solutions.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: A big data-based enterprise information technology service consulting system, the system comprising: Multi-source data acquisition gateways are used to collect heterogeneous data from configuration management databases, monitoring platforms, log platforms, work order platforms, network topology platforms, and external technical benchmark data sources in enterprise information technology environments. They also attach source identifiers, valid time intervals, acquisition costs, data transmission volume, and disturbance levels to the acquired system to the acquisition records. The differential state graph engine is used to map the multi-source heterogeneous data into temporal nodes or temporal edges representing information technology components, component attributes, component dependencies and runtime events, and maintain a baseline state graph and an incremental subgraph relative to the baseline state graph. It establishes conflict edges for conflicting records and calculates the record credibility, and loads the activity subgraph related to the current consultation request into a high-speed memory. A consultation task subgraph generator is used to decompose the consultation request into technical goal nodes, technical constraint nodes, observable indicator nodes, and evidence requirement nodes, and associate the decomposed nodes with the activity subgraph to form a consultation task subgraph. An adaptive evidence collection scheduler is used to calculate the collection value for missing evidence items in the consultation task subgraph according to the ratio between the expected reduction in uncertainty of the candidate technical solution caused by the missing evidence item and the comprehensive cost determined by the collection cost, the amount of data transmitted and the disturbance level. It selects missing evidence items whose collection value meets the preset conditions and controls the multi-source data collection gateway to perform supplementary collection according to the limited data source, time window and sampling rate. The candidate solution generator is used to generate candidate technical solutions based on the updated consultation task subgraph, which apply to component configuration, resource allocation, network routing, caching strategy, data indexing, service deployment or fault recovery strategy in the enterprise information technology environment, and to establish an evidence path for each candidate technical solution consisting of original collection records, temporal nodes or temporal edges and reasoning relationships. The sandbox verification engine is used to reconstruct the local digital twin environment corresponding to the components affected by the candidate technical solutions based on the consultation task subgraph, execute the candidate technical solutions in the isolated sandbox and collect simulation operation indicators, and eliminate candidate technical solutions that do not meet the constraints based on the technical constraint nodes and the observable indicator nodes. A resource orchestrator is used to select between a single-node incremental computing mode and a distributed parallel computing mode based on the number of nodes in the active subgraph and the change density of the incremental subgraph, and to reuse the node features that have not changed in the baseline state graph during the computing process. The consultation output and feedback module is used to output the technical change steps, evidence paths, confidence levels, predicted operating indicators, risk boundaries, and rollback plans corresponding to the candidate technical solutions retained by the sandbox verification engine, and to collect actual operating indicators after the candidate technical solutions are implemented to update the record confidence and candidate solution generation parameters.
[0008] Preferably, the multi-source heterogeneous data includes at least two of the following: processor utilization, memory utilization, storage input / output latency, network latency, packet loss rate, service call chain, application logs, slow database queries, fault alarms, change records, service level targets, IT component versions, and component dependencies.
[0009] Preferably, the differential state graph engine calculates the record credibility for multiple conflicting records of the same component attribute based on data source credibility weight, time freshness, cross-source consistency, and conflict degree, and marks records with credibility below a threshold as records to be verified. The adaptive evidence collection scheduler prioritizes missing evidence items that can verify the record to be verified.
[0010] Preferably, the differential state graph engine associates and stores each incremental subgraph with its corresponding version identifier, start validity time, and end validity time, generates incremental records only for newly added, deleted, or attribute-changed nodes, and uses content hashing to reference the unchanged data blocks in the baseline state graph.
[0011] Preferably, the collected value Determined according to the following relationship: This allows the system to quantitatively weigh the benefits of data collection against production risks; in, For the first One missing piece of evidence. The results of the operational indicators for the candidate technical solutions. For the current set of evidence, As a measure of uncertainty, To reduce data collection costs, To estimate the amount of data to be transmitted, Disturbance level, , and The weights are non-negative.
[0012] Preferably, the sandbox verification engine extracts affected components, upstream dependent components, downstream dependent components, historical load distribution, and fault injection conditions from the consultation task subgraph to generate a local digital twin environment; In the local digital twin environment, baseline playback, candidate technology implementation, peak load testing, and fault recovery testing are performed sequentially, and end-to-end latency, throughput, availability, resource consumption, and recovery time are output.
[0013] Preferably, the candidate solution generator determines the score of the candidate technical solution based on the completeness of evidence, the improvement of sandbox simulation indicators, risk value and resource consumption, and retains at least two alternative solutions for candidate technical solutions with the same technical objectives but whose evidence paths do not intersect.
[0014] Preferably, the consultation output and feedback module calculates the deviation between the actual operating indicators and the predicted operating indicators. When the deviation continuously exceeds the drift threshold, the credibility weight of the data source in the corresponding evidence path is reduced, triggering the adaptive evidence collection scheduler to re-collect and incrementally calibrate the candidate scheme generation parameters.
[0015] The beneficial effects of this invention are as follows: This invention utilizes differential storage and feature reuse of baseline state graphs and incremental subgraphs to reduce the transmission, storage, and computational loads required for full synchronization of multi-source data, graph reconstruction, and repeated inference. It employs adaptive evidence collection based on anticipated uncertainty to reduce quantity and overall cost, improving the integrity of critical data while controlling production system disturbances. It achieves source tracing and integrity verification of consultation conclusions through evidence paths and content hashing. It reduces the risk of direct trial-and-error in the production environment by verifying the impact of technical changes on operational indicators through a local digital twin sandbox. Finally, it continuously calibrates data credibility and solution parameters based on feedback from actual indicators after deployment. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the overall system architecture of Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the differential state diagram construction and calculation process in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the adaptive evidence collection and sandbox verification process in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the closed-loop consultation and model calibration process in Embodiment 1 of the present invention; Figure 5 This is a flowchart illustrating the steps of the enterprise information technology service consulting method based on big data in Embodiment 2 of the present invention.
[0017] In the diagram: 101, Multi-source data acquisition gateway; 102, Differential state graph engine; 103, Consultation task subgraph generator; 104, Adaptive evidence acquisition scheduler; 105, Candidate solution generator; 106, Sandbox verification engine; 107, Resource orchestrator; 108, Consultation output and feedback module. Detailed Implementation
[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0019] Example 1, referring to Figures 1-4A big data-based enterprise information technology service consulting system is deployed in an enterprise private cloud, hybrid cloud, or trusted computing environment. The system includes a multi-source data acquisition gateway 101, a differential state graph engine 102, a consulting task subgraph generator 103, an adaptive evidence acquisition scheduler 104, a candidate solution generator 105, a sandbox verification engine 106, a resource orchestrator 107, and a consulting output and feedback module 108.
[0020] The multi-source data acquisition gateway 101 includes multiple data connectors. These connectors can communicate with the data source via application programming interfaces (APIs), message queues, read-only database connections, telemetry agents, or log forwarding protocols. Each connector maintains a whitelist of data acquisition data, a maximum sampling frequency, data anonymization rules, and a production system disturbance limit. Acquisition records can represent: ; in For source identification, For object identification, For attribute type, For attribute values, For the time of data collection, For the effective time interval, To reduce data collection costs, To estimate the amount of data to be transmitted, Disturbance level, This is used to record the hash of the content.
[0021] The differential state graph engine 102 performs entity alignment for the same component across different data sources based on object identifiers, component fingerprints, and topological relationships. Component fingerprints can be obtained from two or more combinations of component type, network address, instance identifier, certificate fingerprint, deployment namespace, and host identifier. After entity alignment, node attributes and relationship edges are generated, and the mapping from the original records to graph elements is preserved. The record confidence level Q can be calculated using the following formula: ; in For normalization function, As a weight for the credibility of the data source, For freshness of time, For cross-source consistency, For the degree of conflict, to These are weights that are preset or learned through historical feedback. Time freshness can be determined based on the distance between the current time and the end time of the valid interval.
[0022] Baseline state diagram Save the most recent stable snapshot. Incremental subgraph Only newly added nodes, newly added edges, deletion markers, and attribute changes from the previous version to the current version are saved. Unchanged data blocks are referenced via content hashes. Or data from a previous incremental subgraph. When querying the state at a specific point in time, merge data in version order. And the corresponding incremental subgraph.
[0023] To reduce memory usage, the system stores cold nodes unrelated to the current consultation request in a capacity-based storage and loads the active subgraph into high-speed storage. The boundary of the active subgraph can start from the target service node, extend along the call relationship, deployment relationship, and network connection relationship by a preset number of hops, and include nodes associated with constraint nodes or indicator nodes.
[0024] Among them, the resource orchestrator 107 calculates the number of nodes in the active subgraph. and change density Change in density This can be expressed as the ratio of the number of elements in the changing subgraph to the total number of elements in the active subgraph. When... and When the change occurs, a single-node incremental mode is adopted, and only the graph features of the changed nodes and their neighborhoods are recalculated; otherwise, the active subgraph is divided into multiple partitions according to weak connection edges or service boundaries, distributed to multiple processing nodes for parallel computation, and then aggregated at the boundary nodes.
[0025] The consultation task subgraph generator 103 receives consultation requests. Requests can be input by the user or automatically triggered by the monitoring system when metrics are abnormal. The generator extracts the target service, target metric, target threshold, change time window, allowed resources, prohibited change components, and rollback time requirements from the request, and forms task nodes and constraint edges.
[0026] For example, a consultation request might be: "Reduce the 95th percentile response latency of the order interface to less than 200 milliseconds without downtime or adding database instances." The system generates the following nodes: Technical target node: "Order interface P95 latency < 200ms"; Constraint nodes: "No downtime" and "No addition of database instances"; Indicator nodes: "Interface latency, database CPU, slow query count, cache hit rate"; Evidence requirement nodes: "Call chain, slow query samples, index status, cache configuration, and peak load."
[0027] Among them, the adaptive evidence collection scheduler 104 first determines the current evidence set. Calculate the results of each candidate scheme Uncertainty For missing evidence items Estimated collection Then, the expected value of uncertainty And calculate the collection value according to the following formula; ; in, , and It is used to balance acquisition costs, transmission volume, and production disruptions.
[0028] when When the threshold is reached, the scheduler sends a data acquisition instruction to the corresponding connector. The instruction includes at least the data object, time window, sampling rate, maximum data volume, de-identification rules, and timeout conditions. When the production system load exceeds the safety threshold, the connector reduces the sampling rate or suspends data acquisition. After data acquisition is completed, the differential state graph engine 102 generates a new incremental subgraph and updates the task subgraph.
[0029] The candidate solution generator 105 can employ a combination of a rule engine, a graph search model, a machine learning model, or a large language model with a constraint solver. Regardless of the model used, the generator must convert the output into a structured technical change object, which includes target components, preconditions, actions to be performed, expected changes in indicators, risk conditions, verification commands, and rollback actions.
[0030] The evidence path for candidate solutions traces back from the technical change object to the indicator nodes and evidence nodes in the task subgraph, and then back to the original data collection records. Each record in the evidence path retains its source identifier, valid time range, and content hash. Before output, the system verifies that the record is still within the valid range and performs hash verification; if verification fails, the output of the corresponding solution is paused and supplementary data collection is triggered. The candidate solution score S can be represented as: ; in For the completeness of evidence, The improvement amount for sandbox simulation indicators. This is the risk value. For resource consumption, to The weights are used to determine the weights. The system can simultaneously retain multiple alternative solutions with different evidence paths to reduce the impact of bias caused by a single data source.
[0031] The sandbox validation engine 106 determines the affected scope based on the target components of candidate solutions, replicates necessary configurations, service images, data patterns, dependency interface simulators, and historical load models, and builds a local digital twin environment. For external systems that cannot be replicated, service stubs with the same interface protocol and latency distribution are used. The sandbox is isolated from the production network, and write operations only affect sandbox data.
[0032] The verification process may include: first, replaying the baseline load and recording baseline metrics; second, implementing candidate technology changes; third, replaying normal load, peak load, and abnormal load; fourth, injecting preset faults and performing recovery tests; and fifth, comparing simulation metrics with constraint thresholds. If any hard constraint is violated, the candidate solution is eliminated; if a soft constraint is violated, the risk value is increased.
[0033] The consultation output and feedback module 108 generates consultation reports or machine-readable execution plans. The reports include solution prioritization, technical change steps for each solution, evidence paths, confidence levels, prediction indicator ranges, risk boundaries, pre-emptive checks, monitoring indicators, and rollback plans. Machine-readable execution plans can be sent to configuration management, container orchestration, or automated operations and maintenance platforms after manual approval.
[0034] After the scheme is implemented, module 108 collects the actual operating index A within a preset observation period and compares it with the predicted operating index P. The deviation D can be represented by normalized absolute error or quantile error. When D continuously exceeds the drift threshold, the system reduces the credibility weight of the relevant data source or inference relationship, re-triggers evidence collection, and incrementally calibrates the parameters for generating candidate schemes.
[0035] When encountering a database performance consultation scenario, the system observed increased response latency at the enterprise order interface during peak periods. The system determined from the call chain that the bottleneck was concentrated in the database query node, but current evidence lacked slow query text and index statistics. The scheduler calculated the collection value of slow query and index statistics items, limiting collection to read-only mode within ten minutes before and after the peak. The system generated candidate solutions such as "adding a composite index and adjusting the connection pool limit" and "retaining existing indexes and increasing cache expiration time," replaying peak loads in a sandbox. If the first solution reduced P95 latency to the target value without increasing instances and without causing write latency to exceed constraints, then the solution and its index rollback script were output.
[0036] When encountering container resource consultation scenarios, where the target service experiences processor throttling and memory jitter, the system constructs a task subgraph based on container metrics, node resources, and service call graphs. This generates candidate solutions for adjusting request and limit values, changing the number of replicas, and adjusting the thread pool. A sandbox is used to perform peak traffic and single-node failure tests, eliminating solutions with insufficient availability during node failures and outputting resource configurations that meet availability and cost constraints.
[0037] When encountering network fault recovery consultation scenarios, the system addresses the following: the target service is deployed across two data centers, historical work orders show link jitter but insufficient network telemetry sampling. The scheduler only temporarily increases the sampling rate for the relevant links, and the differential state graph records link state changes. The system generates routing weight adjustment and fault handover threshold adjustment schemes, injects packet loss and high latency into a sandbox, and filters and outputs schemes based on the recovery time target.
[0038] In addition, the collaborative work between the differential state graph engine 102 and the adaptive evidence collection scheduler 104 solves the problem of full synchronization and repetitive calculation overhead caused by the continuous changes of multi-source heterogeneous data in the enterprise IT environment when some existing digital twin verification schemes use full network snapshot modeling and intent verification as the path when verifying configuration changes for communication networks. At the same time, it also solves the problem of evidence collection itself causing disturbance to the production system.
[0039] Example 2, as Figure 5 As shown in the figure, this embodiment discloses a method for enterprise information technology service consulting based on big data, which includes the following steps: S1 collects multi-source heterogeneous data from the enterprise's information technology environment and adds source identifiers, valid time intervals, collection costs, data transmission volume, and disturbance levels to the collection records. S2 maps multi-source heterogeneous data into temporal nodes or temporal edges to form a baseline state graph and an incremental subgraph. It establishes conflict edges for conflict records and calculates the record credibility. S3 decomposes the consultation request into technical objectives, technical constraints, observable indicators, and evidence requirements, and associates them with the activity subgraph to form a consultation task subgraph. S4. For missing evidence items, supplementary collection items are selected based on the ratio between the expected reduction in uncertainty of candidate technical solutions and the overall cost, and supplementary collection is carried out according to the limited data source, time window and sampling rate. S5, Generate candidate technical solutions with evidence paths based on the updated consultation task subgraph; S6. Reconstruct the local digital twin environment based on the consultation task subgraph, execute candidate technical solutions in the isolation sandbox, and eliminate candidate technical solutions that do not meet the technical constraints based on the simulation operation indicators. S7. Select either single-node incremental computing mode or distributed parallel computing mode based on the number of nodes in the active subgraph and the change density of the incremental subgraph, and reuse the unchanged node features. S8 outputs the technical change steps, evidence path, confidence level, predicted operating indicators, risk boundary and rollback plan of the retained candidate technical solutions. After the candidate technical solutions are implemented, the actual operating indicators are collected and the confidence level and candidate solution generation parameters are updated and recorded.
[0040] In S3, the formation of the consultation task subgraph includes: parsing the target information technology service, target operation indicators, allowed change time window, available computing resources, service level targets, and prohibited change components in the consultation request; and extending the preset number of hops along the call relationship, deployment relationship, and network connection relationship in the baseline state graph to determine the boundary of the activity subgraph.
[0041] In S7, the selection of single-node incremental computing mode or distributed parallel computing mode includes: calculating the number of nodes N and change density ρ of the active subgraph; when N is less than the first threshold and ρ is less than the second threshold, loading the incremental subgraph on a single node and calling the cached unchanged node features; otherwise, dividing the active subgraph into multiple computing partitions according to component dependencies, distributing them to multiple processing nodes, and synchronously aggregating the results at the boundary nodes.
[0042] In S8, each original collection record in the evidence path is associated with a content hash, source identifier, and valid time interval. When outputting candidate technical solutions, the integrity of the evidence path is checked, and the output of the corresponding candidate technical solution is paused when any original collection record exceeds the valid time interval or the content hash check fails.
[0043] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A big data-based enterprise information technology service consulting system, characterized in that, The system includes: The multi-source data acquisition gateway (101) is used to collect multi-source heterogeneous data from configuration management databases, monitoring platforms, log platforms, work order platforms, network topology platforms and external technical benchmark data sources in the enterprise information technology environment, and to attach source identifiers, valid time intervals, acquisition costs, data transmission volume and disturbance level to the system being collected to the acquisition records. The differential state graph engine (102) is used to map the multi-source heterogeneous data into temporal nodes or temporal edges representing information technology components, component attributes, component dependencies and runtime events, and maintain a baseline state graph and an incremental subgraph relative to the baseline state graph, establish conflict edges for conflicting records and calculate the record credibility, and load the activity subgraph related to the current consultation request into a high-speed memory. Consultation task subgraph generator (103) is used to decompose the consultation request into technical target nodes, technical constraint nodes, observable indicator nodes and evidence requirement nodes, and associate the decomposed nodes with the activity subgraph to form a consultation task subgraph; An adaptive evidence collection scheduler (104) is used to calculate the collection value for missing evidence items in the consultation task subgraph according to the ratio between the expected reduction in uncertainty of the candidate technical solution caused by the missing evidence item and the comprehensive cost determined by the collection cost, the amount of data transmitted and the disturbance level, select missing evidence items whose collection value meets the preset conditions, and control the multi-source data collection gateway (101) to perform supplementary collection according to the limited data source, time window and sampling rate. The candidate solution generator (105) is used to generate candidate technical solutions for component configuration, resource allocation, network routing, caching strategy, data indexing, service deployment or fault recovery strategy in the enterprise information technology environment based on the updated consultation task subgraph, and to establish an evidence path for each candidate technical solution consisting of original collection records, temporal nodes or temporal edges and reasoning relationships. The sandbox verification engine (106) is used to reconstruct the local digital twin environment corresponding to the components affected by the candidate technical solutions based on the consultation task subgraph, execute the candidate technical solutions in the isolated sandbox and collect simulation operation indicators, and eliminate candidate technical solutions that do not meet the constraints based on the technical constraint nodes and the observable indicator nodes. The resource orchestrator (107) is used to select between a single-node incremental computing mode and a distributed parallel computing mode based on the number of nodes in the active subgraph and the change density of the incremental subgraph, and to reuse the node features that have not changed in the baseline state graph during the computing process. The consultation output and feedback module (108) is used to output the technical change steps, evidence paths, confidence levels, predicted operating indicators, risk boundaries and rollback plans corresponding to the candidate technical solutions retained by the sandbox verification engine (106), and to collect actual operating indicators after the candidate technical solutions are implemented to update the record confidence and candidate solution generation parameters.
2. The enterprise information technology service consulting system based on big data according to claim 1, characterized in that, The multi-source heterogeneous data includes at least two of the following: processor utilization, memory utilization, storage input / output latency, network latency, packet loss rate, service call chain, application logs, slow database queries, fault alarms, change records, service level targets, IT component versions, and component dependencies.
3. The enterprise information technology service consulting system based on big data according to claim 1, characterized in that, The differential state graph engine (102) calculates the credibility of multiple conflicting records of the same component attribute based on the data source credibility weight, time freshness, cross-source consistency and conflict degree, and marks the records with credibility below the threshold as records to be verified. The adaptive evidence collection scheduler (104) prioritizes the missing evidence items that can verify the record to be verified.
4. The enterprise information technology service consulting system based on big data according to claim 1, characterized in that, The differential state graph engine (102) stores each incremental subgraph in association with its corresponding version identifier, start validity time, and end validity time. It generates incremental records only for nodes that are added, deleted, or have changed attributes, and uses content hashing to reference the unchanged data blocks in the baseline state graph.
5. The enterprise information technology service consulting system based on big data according to claim 1, characterized in that, The value of the collection Determined according to the following relationship: ; in, For the first One missing piece of evidence. The results of the operational indicators for the candidate technical solutions. For the current set of evidence, As a measure of uncertainty, To reduce data collection costs, To estimate the amount of data to be transmitted, Disturbance level, , and The weights are non-negative.
6. The enterprise information technology service consulting system based on big data according to claim 1, characterized in that, The sandbox verification engine (106) extracts affected components, upstream dependent components, downstream dependent components, historical load distribution and fault injection conditions from the consultation task subgraph to generate a local digital twin environment; In the local digital twin environment, baseline playback, candidate technology implementation, peak load testing, and fault recovery testing are performed sequentially, and end-to-end latency, throughput, availability, resource consumption, and recovery time are output.
7. The enterprise information technology service consulting system based on big data according to claim 1, characterized in that, The candidate solution generator (105) determines the score of the candidate technical solution based on the completeness of evidence, the improvement of sandbox simulation indicators, risk value and resource consumption, and retains at least two alternative solutions for candidate technical solutions with the same technical objectives but whose evidence paths do not intersect.
8. The enterprise information technology service consulting system based on big data according to claim 1, characterized in that, The consultation output and feedback module (108) calculates the deviation between the actual operating indicators and the predicted operating indicators. When the deviation exceeds the drift threshold continuously, it reduces the credibility weight of the data source in the corresponding evidence path, triggers the adaptive evidence collection scheduler (104) to collect again, and performs incremental calibration on the candidate scheme generation parameters.
9. A big data-based enterprise information technology service consulting method, applied to the system described in any one of claims 1-8, characterized in that, Includes the following steps: S1 collects multi-source heterogeneous data from the enterprise's information technology environment and adds source identifiers, valid time intervals, collection costs, data transmission volume, and disturbance levels to the collection records. S2 maps multi-source heterogeneous data into temporal nodes or temporal edges to form a baseline state graph and an incremental subgraph. It establishes conflict edges for conflict records and calculates the record credibility. S3 decomposes the consultation request into technical objectives, technical constraints, observable indicators, and evidence requirements, and associates them with the activity subgraph to form a consultation task subgraph. S4. For missing evidence items, supplementary collection items are selected based on the ratio between the expected reduction in uncertainty of candidate technical solutions and the overall cost, and supplementary collection is carried out according to the limited data source, time window and sampling rate. S5, Generate candidate technical solutions with evidence paths based on the updated consultation task subgraph; S6. Reconstruct the local digital twin environment based on the consultation task subgraph, execute candidate technical solutions in the isolation sandbox, and eliminate candidate technical solutions that do not meet the technical constraints based on the simulation operation indicators. S7. Select either single-node incremental computing mode or distributed parallel computing mode based on the number of nodes in the active subgraph and the change density of the incremental subgraph, and reuse the unchanged node features. S8 outputs the technical change steps, evidence path, confidence level, predicted operating indicators, risk boundary and rollback plan of the retained candidate technical solutions. After the candidate technical solutions are implemented, the actual operating indicators are collected and the confidence level and candidate solution generation parameters are updated and recorded.
