Search

CN-121530964-B - Large file transmission state monitoring method based on multi-service collaboration

CN121530964BCN 121530964 BCN121530964 BCN 121530964BCN-121530964-B

Abstract

The invention discloses a large file transmission state monitoring method based on multi-service cooperation, which relates to the technical field of computer networks and data transmission, and the invention firstly completes the consistency of time reference and statistical caliber to form a comparable transmission stream record by uniformly collecting key telemetry such as effective throughput, time delay, packet loss and the like on a branch side and superposing service priority, time effect and region compliance strategies, thereby solving the problem of misjudgment caused by different caliber and time sequence mismatch of cross-domain monitoring data; on the basis, a multi-service collaborative monitoring model is adopted, three types of constraints including performance, SLA and compliance are fused into a single health measurement, and continuous penalty terms are brought into the cross-region risk, so that a monitoring result can reflect performance degradation and hidden danger of compliance at the same time, and distortion of a compliance path is avoided being ignored by only looking at throughput.

Inventors

  • ZHANG YUBING
  • CHEN DEGUI

Assignees

  • 北京英创思信息技术有限公司

Dates

Publication Date
20260512
Application Date
20251124

Claims (9)

  1. 1. The large file transmission state monitoring method based on multi-service cooperation is characterized by comprising the following steps of: Step S1, deploying monitoring agents at a plurality of branch nodes, and collecting local transmission telemetry and service strategies, wherein the transmission telemetry comprises effective throughput, round trip delay, packet loss/retransmission rate, error codes, queue length and bandwidth utilization rate, and the service strategies comprise service priority, data timeliness requirements and region compliance strategies; step S2, carrying out time synchronization and caliber consistency on the acquired data to generate a transport stream record with a region and compliance label; Step S3, constructing a multi-service collaborative monitoring model based on the transport stream records, and outputting a composite health degree score and a compliance risk level for each service stream; Step S4, generating a monitoring result according to the score and the risk level, wherein the monitoring result comprises prioritized alarms, root cause indication and scheduling suggestions, and the prioritized alarms, the root cause indication and the scheduling suggestions are provided for an upper layer system through a northbound interface; Step S5, online updating of model parameters and thresholds is carried out according to false alarm/missing alarm statistics of historical alarms; wherein the scheduling suggestions include, but are not limited to, speed limits, path isolation, or preferred paths within the region; The multi-service collaborative monitoring model carries out weighted fusion on the performance index, the SLA constraint and the compliance constraint to obtain the composite health score, and the composite health score is used for sequencing service flows and comparing the service flows with an alarm threshold value to trigger alarm and suggestion; The step of weighted fusion of the composite health score comprises the following steps: Step S31, unifying the effective throughput, round trip time RTT, packet loss/retransmission rate, bandwidth utilization rate, direction attribute and statistical caliber of completion time dividing value, determining the business target value and warning limit of each index, and using the sampling window and dividing statistics as the base line of normalization and penalty item in accordance with SLA definition; step S32, in the scoring link, performing interval linear mapping on each original index and cutting off the two ends of each original index into utility scores: , Wherein, the Representing traffic flow index At index The normalization effect on the utility of the above is that, Representing traffic flows At the index Is used for the original observation of (a), Is taken as an index Is set up in the course of the warning limit of (2), Is taken as an index Is used for the service target value of (1), As the direction factor, the direction factor is used, The larger the representation the better, The smaller the representation the better, Representation of real numbers Cutting to interval ; Step S33, combining the static weight and the learning weight to obtain a final weight: , Wherein, the Index for index Is used to determine the final weight of the (c) for the (c), In order for the coefficient of fusion to be a function of, For the static configuration of the weights, In order to learn the resulting weight(s), And (3) with Identifying a corner mark for a source; Step S34, carrying out dimensionless aggregation on the negative deviation exceeding the target: , Wherein, the Indexing traffic streams An amount of SLA breach polymerization, In order to incorporate the set of metrics for scoring, As an absolute value of the absolute value, Corner mark for positive operator ; Step S35, mapping the discrete handover risk as Successive penalty terms of (2): , Wherein, the Indexing traffic streams Is used for the compliance penalty term of (c), In order to be a level of risk of a compliant handoff, Constant angle sign for risk level upper bound ; Step S36, weighting and summing the utility, deducting two penalty items and cutting Obtaining the final health degree: , Wherein, the Indexing traffic streams Is marked with a final composite health degree of (2) To deduct the penalized and cut state corner marks, 、 For the index weight and the normalized utility, For the SLA penalty term factor, For the amount of polymerization of SLA violations, For the compliance penalty term coefficients, Is a compliance penalty term; Step S37, layering according to service priority classes, and constructing a multi-level alarm threshold value by using the hierarchical points of the historical health sample: , Wherein, the Indexing for priority class Index at alarm level A threshold of health at the point of the network, As a quantile function, the quantile parameter is , Corner mark for compound health degree Indicating category layering, corner mark Indicating the alert level.
  2. 2. The method for monitoring the transmission state of large files based on multi-service collaboration as claimed in claim 1, wherein the region compliance policy comprises a limitation of regions where transmission paths pass, a limitation of data storage/processing places and a cross-border transmission condition, and the monitoring agent attaches a region identifier and a processing position identifier to each transmission section.
  3. 3. The method for monitoring the transmission state of a large file based on multi-service collaboration according to claim 1, wherein the monitoring model comprises a graph neural network, and the branch nodes are used as graph nodes and the transmission links are used as graph edges for modeling; the node characteristic codes a service policy vector, and the edge characteristic codes a link state, and is used for estimating the path-level degradation probability and the influence range.
  4. 4. The method for monitoring the transmission state of a large file based on multi-service collaboration as claimed in claim 1, wherein the method comprises the step of checking a compliance path: and comparing autonomous domain, region and medium point information of the real-time path with the region compliance policy, marking the risk of the cross region and merging the risk into the compliance risk level.
  5. 5. The method for monitoring the transmission state of large files based on multi-service collaboration according to claim 1, wherein the monitoring result comprises a prioritized alarm queue, and an executable treatment suggestion type and suggestion effective range are given for each alarm, and the network configuration is not directly changed.
  6. 6. The method for monitoring the transmission state of large files based on multi-service collaboration according to claim 1, wherein the online updating comprises a threshold self-adaption module based on reinforcement learning, and the alarm threshold and each index weight are adjusted according to the false alarm/missing report statistics of the historical alarms and the critical service SLA satisfaction rate so as to reduce the comprehensive cost and improve the SLA satisfaction rate; the reinforcement learning reward function is constructed as follows: step S51, setting off-line playback or weak supervision marking caliber according to time steps Counting false alarm, missing report and key business SLA achievement conditions in a sliding window, and recording the number of alarm bars and adjacent step variation as input of rewarding each item; Step S52, calculating the weighted cost of false alarm and missing alarm at each time step: , Wherein, the Representing time steps Is used to determine the error cost of (1), In order to false-positive the weight coefficient, In order to miss-report the weight coefficient, In order to be a false positive rate, In order to achieve the rate of missing report, For the number of false alarm bars, For the number of the missed report bars, To evaluate the total sample size, superscript 、 、 The item category corner mark; Step S53, forward rewards are given to the achievement rate of the key business: , , Wherein, the Is a time step The SLA reward item(s), The coefficients are awarded for the SLA, For the critical service SLA to meet the rate, In order to meet the critical traffic sample number of SLAs, For the total number of key business samples, superscript 、 The item category corner mark; Step S54, the scale and the jitter are measured by using the difference between the number of alarm bars and the adjacent steps: , Wherein, the As a result of the regular cost of the algorithm, For the number of regularized coefficients, For the dithering of the regularization coefficient, Is a time step Is used for controlling the number of the alarm bars, For the number of the warning strips in the last step, For scale normalization constants, superscript 、 、 、 The item category corner mark; Step S55, at the same time step, synthesizing the rewards and the costs into scalar returns: , Wherein, the Is a time step Is a real-time reward of (1), The items are awarded for the SLA, In the event of an erroneous cost, Is a regular cost; Step S56, encoding the near-term health distribution, branch load and period label into a state: , Wherein, the Is a time step Is used to determine the state vector of (1), For the composite health degree in the window Is used for the average value of (a), Is the standard deviation of the same window opening, For the branch load index to be the same, Is the one-hot vector of the time slot label, For the period index of time, Is the proportion of a three-division interval square, Step S57, the action is composed of fine tuning steps of threshold and weight, and the projection satisfies the constraint: , , , Wherein, the Is the index weight Is used for the step size of (a), Is a threshold value Is used for the step size of (a), 、 For the amount of projection front-end, For projection onto probability simplex Is a function of the operator of (2), For projection onto a monotonic threshold set Is a function of the operator of (2), For the upper bound of the step size, The next time step angle mark.
  7. 7. The method for monitoring the transmission state of a large file based on multi-service collaboration according to claim 1, wherein the monitoring model is updated by a federal learning framework: Each branch node trains model parameters locally and reports encryption gradients to an aggregation end, and the aggregation end completes model aggregation and consistency verification and then issues updating.
  8. 8. The method for monitoring the transmission state of a large file based on multi-service collaboration according to claim 7, wherein the federal learning introduces a differential privacy mechanism in the aggregation stage, and adds noise to the aggregate.
  9. 9. The method for monitoring the transmission state of a large file based on multi-service collaboration according to claim 1, wherein the time synchronization and caliber consistency comprises: and carrying out clock synchronization, uniform sampling period and index definition on the branch nodes, and calculating effective throughput and retry amplification factors on the monitoring agent side.

Description

Large file transmission state monitoring method based on multi-service collaboration Technical Field The invention relates to the technical field of computer networks and data transmission, in particular to a large file transmission state monitoring method based on multi-service cooperation. Background In the businesses such as financial merging, supply chain reconciliation, compliance audit and the like, the across-country enterprises need to synchronize general files among multiple branches, such as objects needing fragmentation or breakpoint continuous transmission, network conditions of all branches are obvious in difference, high time delay, jitter and bandwidth fluctuation exist in partial areas, meanwhile, the requirements of local data compliance are required to be met, and by taking European Union as an example, GDPR sets compliance conditions, such as compliance determination or standard contract terms, to the across-country transmission of a third country, and data localization requirements exist in partial countries or industries. The existing monitoring scheme carries out health assessment by continuously collecting end-to-end telemetry (throughput, time delay, error codes and the like) and combining an anomaly detection/prediction model, and part of the system adopts federal learning to reduce the outgoing risk of original data or combines a dynamic threshold value to adapt to network fluctuation. In the existing monitoring, the model is shifted along with the rapid change of network and business strategy in the coexistence scene of multi-branch, heterogeneous network and compliance, the false alarm/missing report rate is increased, the region and compliance strategy of the monitoring index are not coded, the cross-area route or the non-compliance link cannot be found in time, the business priority and SLA are not incorporated into the monitoring domain, the executable priority alarm cannot be provided for the key business flow, in addition, the cross-node time reference is inconsistent and the index caliber is not uniform, the predicted result deviates from the actual transmission state, the quality indexes such as the effective throughput, the retransmission rate and the like are ignored by judging the throughput threshold value on the heterogeneous link, and the resource allocation decision is difficult to support, so that a large-file transmission state monitoring method based on multi-business cooperation is needed to solve the problems. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. The invention provides a large file transmission state monitoring method based on multi-service cooperation, which solves the problem that the existing monitoring is difficult to support because the synchronous large files are restricted by compliance and priority under a multi-branch heterogeneous network. In order to solve the technical problems, the invention provides the following technical scheme: The embodiment of the invention provides a large file transmission state monitoring method based on multi-service cooperation, which comprises the following steps: Step S1, deploying monitoring agents at a plurality of branch nodes, and collecting local transmission telemetry and service strategies, wherein the transmission telemetry comprises effective throughput, round trip delay, packet loss/retransmission rate, error codes, queue length and bandwidth utilization rate, and the service strategies comprise service priority, data timeliness requirements and region compliance strategies; step S2, carrying out time synchronization and caliber consistency on the acquired data to generate a transport stream record with a region and compliance label; Step S3, constructing a multi-service collaborative monitoring model based on the transport stream records, and outputting a composite health degree score and a compliance risk level for each service stream; Step S4, generating a monitoring result according to the score and the risk level, wherein the monitoring result comprises prioritized alarms, root cause indication and scheduling suggestions, and the prioritized alarms, the root cause indication and the scheduling suggestions are provided for an upper layer system through a northbound interface; Step S5, online updating of model parameters and thresholds is carried out according to false alarm/missing alarm statistics of historical alarms; where the scheduling suggestions include, but are not limited to, speed limits, path isolation, or preferred paths within the region. The regional compliance strategy comprises limitation of regions where transmission paths pass, limitation of data storage/processing places and cross-border transmission conditions, and the monitoring agent attaches region identification and processing position identification to each transmission paragraph. The multi-service collaborative monitoring model carries out