CN-121998400-A - Flow node processing efficiency monitoring method based on decision flow

Abstract

The invention belongs to the technical field of decision flow node efficiency monitoring, and particularly discloses a flow node processing efficiency monitoring method based on decision flow, which is used for analyzing efficiency rigidity by collecting input data quantity and processing time consumption of decision flow nodes to quantify efficiency vulnerability of the nodes, improving efficiency state judgment accuracy to the maximum extent, simultaneously determining downstream nodes of each efficiency fragile node according to data transfer relation among the nodes on the basis of identifying the efficiency fragile nodes based on the efficiency rigidity, and by performing bidirectional coupling evaluation of forward load conduction from upstream to downstream and reverse back pressure blocking from downstream to upstream on node pairs in the group, efficiency analysis can be deduced from causal logic deep into the bottom layer from time-consuming observation of the surface layer, and a risk conduction path generated by the bidirectional coupling evaluation is provided for an efficiency degradation propagation link for operation and maintenance personnel, so that risk source nodes can be rapidly positioned, and repair time is greatly shortened.

Inventors

• WANG ZIPEI
• YI XUEFEI
• Gan Haipeng

Assignees

• 国投人力资源服务有限公司

Dates

Publication Date

20260508

Application Date

20260401

Claims (10)

1. 1. The flow node processing efficiency monitoring method based on the decision flow is characterized by comprising the following steps of: Acquiring the input data quantity and the time consumption of processing data of each node in the decision flow in real time in a continuous time window, and determining the instantaneous efficiency rigidity of the node according to the change degree of the input data quantity and the change degree of the time consumption of processing data in each window; Comparing the instantaneous efficiency rigidity of each node with an efficiency rigidity reference based on historical data statistics, and identifying an efficiency fragile node; Determining downstream nodes of each efficiency fragile node according to the data flow direction between nodes defined by the decision demand graph to form an upstream-downstream association group; Performing bidirectional efficiency influence coupling analysis on node pairs in an upstream-downstream association group in an upstream-downstream direction and a downstream-upstream direction respectively; based on the bidirectional efficiency influence coupling analysis result, carrying out efficiency coupling deterioration judgment on node pairs in the upstream and downstream association groups; And aggregating the nodes in the efficiency coupling deterioration state according to the data flow direction to form an inefficient conduction path, and generating hierarchical early warning information according to the path length and the node deterioration type.
2. 2. The method for monitoring processing efficiency of a process node based on decision flow according to claim 1, wherein determining the instantaneous efficiency stiffness of the node comprises: counting the total amount of received input data and the average time consumed for processing the data for each node in each sliding time window; Comparing the total input data of the current time window with the previous adjacent time window, calculating the change rate of the input data, comparing the average time consumption of the processing data of the current time window with the previous adjacent time window, and calculating the change rate of the time consumption of the processing; The ratio of the rate of change of the amount of input data to the rate of change of the processing time is defined as the instantaneous efficiency stiffness of the node over the current time window.
3. 3. The method for monitoring processing efficiency of a flow node based on decision flow according to claim 1, wherein the identifying the fragile node is performed by: Collecting node operation logs of the decision flow in a plurality of historical operation periods, and dividing each historical operation period into a plurality of time slices according to the characteristics of business tides; calculating the instantaneous efficiency rigidity of a historical time window by using the operation log of each node in each time segment, and aggregating to form a historical efficiency rigidity data set of the node under the corresponding time segment; Constructing an efficiency rigidity distribution statistical feature aiming at a historical efficiency rigidity data set of each node in each time segment, extracting an efficiency rigidity confidence interval under a preset confidence level, and taking the lower limit value of the confidence interval as an efficiency rigidity benchmark of the node in the corresponding time segment; And acquiring a time segment to which the current time belongs in real time, calling a corresponding efficiency rigidity reference, and judging that a node is an efficiency fragile node if the current instantaneous efficiency rigidity of the node is lower than the reference and the duration exceeds the length of an observation window.
4. 4. The method for monitoring processing efficiency of a flow node based on decision flow according to claim 1, wherein the upstream and downstream association groups comprise the following steps: extracting a direct downstream node of each efficiency fragile node according to an inter-node data flow defined by the decision demand graph; Each efficiency fragile node is combined with the immediate downstream node into an upstream-downstream association group.
5. 5. The method for monitoring processing efficiency of a flow node based on decision flow according to claim 1, wherein the upstream and downstream association groups further comprise the following steps: When a certain efficiency fragile node has a plurality of direct downstream nodes, sorting is carried out according to the data receiving quantity duty ratio of each downstream node from big to small; and pairing the efficiency fragile nodes with each direct downstream node respectively according to the sequencing result to form a plurality of upstream and downstream association groups, and labeling priority identification for each association group.
6. 6. The method for monitoring processing efficiency of a flow node based on decision flow according to claim 1, wherein the bidirectional efficiency impact coupling analysis is implemented as follows: respectively extracting instantaneous efficiency rigidity sequences of upstream nodes and downstream nodes in a plurality of continuous time windows for node pairs in an upstream-downstream association group, and simultaneously recording data flow transfer quantity output from the upstream nodes to the downstream nodes in each time window; Setting the maximum exploration hysteresis step length, and sequentially calculating cross-correlation coefficients between the instantaneous efficiency stiffness sequences of the upstream nodes and the instantaneous efficiency stiffness sequences of the downstream nodes with different hysteresis step lengths; If the cross correlation coefficient under a certain hysteresis step length exceeds a forward coupling threshold value and the physical time corresponding to the hysteresis step length is smaller than the maximum tolerant conduction delay, primarily judging that the forward efficiency coupling influence exists; And in a hysteresis window in which the forward coupling influence is primarily judged, comparing the data flow quantity of the current window with the data flow quantity of the last adjacent window, calculating the data flow descending amplitude, and if the data flow descending amplitude exceeds the flow attenuation limit value, confirming that the forward efficiency coupling influence exists.
7. 7. The method for monitoring processing efficiency of a flow node based on decision flow according to claim 6, wherein the efficiency-affecting coupling analysis in a downstream-to-upstream direction in the bidirectional efficiency-affecting coupling analysis is implemented as follows: Collecting queue backlog length of a downstream node in each time window and blocking waiting time consumption generated by the upstream node due to waiting for the downstream node to release resources for node pairs in an upstream-downstream association group, and taking a normalized product of the queue backlog length and the blocking waiting time consumption as a downstream blocking degree; for an upstream-downstream association group, synchronously recording an upstream node instantaneous efficiency stiffness sequence, a downstream node blocking degree sequence and a local data backlog sequence of an upstream node in a continuous time window; setting the maximum exploration lag step length, and sequentially calculating the cross correlation coefficient between the downstream node blocking degree sequence and the upstream node instantaneous efficiency stiffness sequence of different lag step lengths; If the cross correlation coefficient under a certain hysteresis step length is smaller than the reverse coupling threshold value and the physical time corresponding to the hysteresis step length is smaller than the maximum response time delay, the reverse efficiency coupling influence is primarily judged to exist; And in a hysteresis window in which the reverse coupling influence is primarily determined, comparing the local data backlog quantity of an upstream node of the current window with the last adjacent window, calculating the data backlog rising amplitude, and if the data backlog rising amplitude exceeds the backlog deterioration limit value, determining that the reverse efficiency coupling influence exists.
8. 8. The method for monitoring processing efficiency of a flow node based on decision flow according to claim 7, wherein the determination of efficiency coupling degradation is performed by: If a certain upstream and downstream association group only has the forward efficiency coupling influence, determining that the downstream node is in a passive coupling deteriorated state influenced by upstream conduction; if a certain upstream and downstream association group has both forward efficiency coupling influence and reverse efficiency coupling influence, the upstream and downstream nodes are judged to be in a mutual-aggravated bidirectional coupling worsening state.
9. 9. The method for monitoring processing efficiency of a decision flow based process node of claim 1, wherein said inefficient conductive path comprises the steps of: Taking a node in a passive coupling degradation state or a bidirectional coupling degradation state as a path seed node; Starting from a path seed node, tracing back upstream along the reverse sequence of the data transmission direction, if the adjacent upstream node is in any one of an efficiency weak state, a passive coupling worsening state or a bidirectional coupling worsening state, taking the upstream node into the current path, continuing tracing back upstream until encountering a node which is not judged to be in any worsening state or reaching a decision flow source, and determining the initial boundary of the path; Starting from the path seed node, traversing forward downstream along the sequence of the data transmission direction, if the adjacent downstream node is in a passive coupling deteriorated state or a bidirectional coupling deteriorated state, incorporating the downstream node into the current path, continuing traversing downstream until encountering a node which is not judged to be in any deteriorated state or reaching the end of the decision flow, and determining the termination boundary of the path; And marking a sequence formed by continuous nodes between the starting boundary and the ending boundary as an inefficient conduction path, and recording a starting node identification and a passing node list for the path.
10. 10. The method for monitoring processing efficiency of a flow node based on decision flow according to claim 1, wherein the hierarchical early warning information is generated by the following steps: Counting the total number of nodes contained in each low-efficiency conduction path as path length, and summarizing the deterioration type of each node in the path; If a node in a bidirectional coupling worsening state exists in the path, marking the path as a high risk level; If the path only comprises the node of the passive coupling worsening state and the path length exceeds the length threshold, marking the path as a medium risk level; if the path only comprises the node in the passive coupling degradation state and the path length is lower than the length threshold value, marking the path as a low risk level; And combining the path identification, the path length, the node deterioration type and the risk level to generate hierarchical early warning information.

Description

Flow node processing efficiency monitoring method based on decision flow Technical Field The invention belongs to the technical field of decision flow node efficiency monitoring, and particularly discloses a flow node processing efficiency monitoring method based on decision flow. Background With the penetration of enterprise digital transformation, the business rules, decision nodes and branch nodes are organized through decision flow arrangement, so that the automatic processing of the business flow is realized, and the method becomes an important means for improving the operation efficiency. However, the decision-making process of modern enterprises usually presents high complexity and large-scale characteristics, often including tens to hundreds of heterogeneous decision nodes, and the processing efficiency of the nodes directly determines the response speed of the whole business process. Existing monitoring methods for decision flow node processing efficiency mainly rely on threshold alarms of a single performance index, for example, when the processing time of a node reaches an upper limit, the alarms are triggered. However, the method has the technical defects that firstly, a single index is difficult to comprehensively describe the real health state of the node, the processing time consumption of the node is greatly influenced by the fluctuation of the input data quantity, and only the time consumption is concerned, so that whether the node is slow in response due to the increase of the input data quantity or low in efficiency due to the reduction of the processing capacity of the node cannot be distinguished. For example, when the amount of input data increases sharply, a moderate increase in processing time is a normal phenomenon, but if a fixed threshold alarm is adopted, a large number of false positives are likely to occur, whereas when performance degradation occurs in a node, the processing time may gradually rise under the condition that the amount of data is unchanged, but a single index is difficult to capture such a subtle change, resulting in false negatives. Secondly, the existing monitoring method regards each node as an isolated individual, and correlation analysis on interaction among the nodes is lacking. In the decision flow, there is a data transfer relationship between nodes, and efficiency fluctuation of an upstream node is conducted to a downstream node through data transfer, otherwise, a processing bottleneck of the downstream node may affect the upstream node through back pressure. The traditional single-point monitoring can not identify the efficiency coupling relation among the nodes, so that when risks propagate along the data flow direction, the root node is difficult to position, and timeliness and accuracy of fault investigation are restricted. Disclosure of Invention In order to solve the above technical problems or at least partially solve the above technical problems, the present invention provides a method for monitoring processing efficiency of a flow node based on decision flow. The technical scheme includes that the flow node processing efficiency monitoring method based on the decision flow comprises the following steps of collecting input data quantity and time consumption of processing data of each node in the decision flow in a continuous time window in real time, and determining instantaneous efficiency rigidity of the node according to the change degree of the input data quantity and the change degree of the time consumption of the processing data in each window. And comparing the instantaneous efficiency rigidity of each node with an efficiency rigidity standard based on historical data statistics, and identifying an efficiency fragile node. And determining downstream nodes of each efficiency fragile node according to the inter-node data flow direction defined by the decision demand graph to form an upstream-downstream association group. And respectively carrying out bidirectional efficiency influence coupling analysis on node pairs in the upstream-downstream association group in the upstream-downstream direction and the downstream-upstream direction. And carrying out efficiency coupling deterioration judgment on the node pairs in the upstream and downstream association groups based on the bidirectional efficiency influence coupling analysis result. And aggregating the nodes in the efficiency coupling deterioration state according to the data flow direction to form an inefficient conduction path, and generating hierarchical early warning information according to the path length and the node deterioration type. By combining all the technical schemes, the method has the advantages that 1, efficiency stiffness analysis is carried out by collecting the input data quantity of the decision flow node and the time consumption of processing the data so as to quantify the efficiency vulnerability of the node, the efficiency judgment realizes the dynamic decoupling o