Graph compression method and device

Abstract

The invention discloses a graph compression method and device, and the method comprises the steps: determining a critical path and a non-critical path of an original operation and maintenance topological graph, and the original operation and maintenance topological graph comprises a plurality of nodes of a distributed system and edges representing the connection relation between the nodes; on the premise of keeping the critical path, performing path folding processing on the non-critical path to obtain a compressed topological graph; and displaying the compressed topological graph in response to a display operation instruction.

Description

Technical Field

[0001] This application relates to the field of graph compression technology, and in particular to a graph compression method and apparatus. Background Technology

[0002] Commercial banks' batch distributed systems (referred to as distributed systems) carry core financial business processes. Their topology contains tens of thousands of nodes, and the dependencies between nodes form complex business links. With the continuous expansion of business, the system faces two prominent challenges: First, the massive number of nodes leads to a decrease in the efficiency of topology graph visualization and a significant increase in fault location time; second, traditional compression algorithms are difficult to adapt to the unique characteristics of banking systems, such as unweighted edges and weighted nodes, and lack the ability to perceive pre-calculated business weights, posing a risk of mistakenly folding high-value business nodes. The topological differences between different business lines further highlight the complexity of the problem—for example, the dense long-chain structure in the clearing system and the clustered short paths in the credit system require different compression strategies, and fixed threshold mechanisms are prone to under-compression or over-compression in such scenarios, affecting the efficiency of operation and maintenance response.

[0003] Current mainstream solutions have three limitations: First, compression algorithms based on geometric features (such as the Douglas-Peucker algorithm) mainly focus on topological shape simplification and fail to effectively utilize pre-calculated node weights, which may lead to the incorrect compression of important nodes. Second, static parameter settings cannot adapt to the topological features of different business systems. The same threshold has limited compression effect in sparse systems and may cause loss of effective information in dense systems. Third, the folding operation is irreversible and does not retain the business attributes of intermediate nodes, which increases the difficulty of fault analysis and makes it difficult to meet the requirements of traceability of the operation process in the financial field.

[0004] In other words, existing graph compression methods may remove some important information from the topology graph when compressing the topology graph of a distributed system, making it impossible for users to view this important information from the compressed topology graph, i.e., there is a problem of information loss. Summary of the Invention

[0005] Therefore, this application discloses the following technical solution:

[0006] The first aspect of this application provides a graph compression method, comprising:

[0007] Determine the critical and non-critical paths of the original operation and maintenance topology graph, which includes multiple nodes of the distributed system and edges representing the connection relationships between the nodes;

[0008] While maintaining the critical path, the non-critical paths are folded to obtain a compressed topology graph;

[0009] In response to a display operation command, the compressed topology map is displayed.

[0010] Optionally, the step of performing path folding on the non-critical paths to obtain a compressed topology map includes:

[0011] In the non-critical paths, select low-weight long paths whose length is greater than the length threshold and whose average weight value is less than the weight threshold.

[0012] The low-weight long paths are folded to obtain a compressed topology graph.

[0013] Optionally, the step of performing path folding on the low-weight long paths to obtain a compressed topology graph includes:

[0014] From the node set of the original operation and maintenance topology, delete all intermediate nodes of the low-weight long path. Intermediate nodes include nodes in the low-weight long path other than the first and last nodes.

[0015] From the edge set of the original operation and maintenance topology graph, delete all edges contained in the low-weight long path, and add shortcut edges that directly connect the first and last nodes of the low-weight long path to obtain the updated edge set; the shortcut edges have shortcut edge weights, which are determined based on the comprehensive weights of the intermediate nodes of the low-weight long path.

[0016] The node set after deleting all intermediate nodes of the low-weight long path and the updated edge set constitute the compressed topology graph.

[0017] Optionally, before performing path folding on the non-critical paths to obtain the compressed topology map, the method further includes:

[0018] To achieve the preset desired compression ratio;

[0019] The initial length threshold and the initial weight threshold are iteratively updated based on the desired compression ratio to obtain an optimal threshold pair, which includes a length threshold and a weight threshold for filtering low-weight long paths.

[0020] Optionally, displaying the compressed topology map includes:

[0021] Display the compressed topology graph and label the minimum hop count of the critical path in the compressed topology graph.

[0022] A second aspect of this application provides a graph compression apparatus, comprising:

[0023] A determining unit is used to determine the critical and non-critical paths of the original operation and maintenance topology graph, which includes multiple nodes of the distributed system and edges representing the connection relationships between the nodes;

[0024] The folding unit is used to fold the non-critical paths while maintaining the critical paths, so as to obtain a compressed topology map.

[0025] The display unit is used to display the compressed topology map in response to a display operation command.

[0026] Optionally, the folding unit performs path folding processing on the non-critical paths to obtain a compressed topology map, including:

[0027] In the non-critical paths, select low-weight long paths whose length is greater than the length threshold and whose average weight value is less than the weight threshold.

[0028] The low-weight long paths are folded to obtain a compressed topology graph.

[0029] Optionally, the folding unit performs path folding processing on the low-weight long paths to obtain a compressed topology graph, including:

[0030] From the node set of the original operation and maintenance topology, delete all intermediate nodes of the low-weight long path. Intermediate nodes include nodes in the low-weight long path other than the first and last nodes.

[0031] From the edge set of the original operation and maintenance topology graph, delete all edges contained in the low-weight long path, and add shortcut edges that directly connect the first and last nodes of the low-weight long path to obtain the updated edge set; the shortcut edges have shortcut edge weights, which are determined based on the comprehensive weights of the intermediate nodes of the low-weight long path.

[0032] The node set after deleting all intermediate nodes of the low-weight long path and the updated edge set constitute the compressed topology graph.

[0033] Optionally, before the folding unit performs path folding processing on the non-critical paths to obtain the compressed topology map, it is further used for:

[0034] To achieve the preset desired compression ratio;

[0035] The initial length threshold and the initial weight threshold are iteratively updated based on the desired compression ratio to obtain an optimal threshold pair, which includes a length threshold and a weight threshold for filtering low-weight long paths.

[0036] Optionally, the display unit displays the compressed topology map, including:

[0037] Display the compressed topology graph and label the minimum hop count of the critical path in the compressed topology graph.

[0038] The beneficial effects of this solution are as follows: Before compressing the original operation and maintenance topology map, critical and non-critical paths are pre-determined, and path folding is performed only on non-critical paths while the critical paths remain unchanged. This not only compresses the original operation and maintenance topology map through path folding, reducing its complexity, but also avoids the loss of important information caused by folding critical paths. Thus, the problem of information loss is avoided while making it convenient for users to view the topology map. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0040] Figure 1 This is a flowchart of a graph compression method provided in an embodiment of this application;

[0041] Figure 2 This is a schematic diagram illustrating the implementation principle of an adaptive algorithm for dynamic thresholds provided in an embodiment of this application;

[0042] Figure 3 This is a flowchart of another graph compression method provided in an embodiment of this application;

[0043] Figure 4 This is a schematic diagram of the structure of a compression device provided in an embodiment of this application. Detailed Implementation

[0044] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0045] To better understand the technical solution of this application, some terms that may be involved in this application will be explained first.

[0046] Batch processing refers to a technical model that processes a large number of similar tasks or data at once.

[0047] A batch system refers to a system that connects to and uses a scheduling platform to execute batch jobs.

[0048] Batch jobs or batch tasks refer to an executable unit, such as a program or script, used to process a batch of data at once.

[0049] A batch node or node is a scheduling configuration unit for batch jobs, which typically includes batch tasks, scheduling time, execution parameters, and other key configurations.

[0050] Child nodes are simplified copies of a node's configuration and depend on the node's existence. They can only be configured with a small amount of information, such as execution parameters; most of their configuration is inherited from the node's configuration. They are mainly used for handling data partitioning and improving concurrency capabilities.

[0051] In this embodiment, a node can be considered as a container instance in a batch system (also known as a distributed batch system or a distributed system) used to process batch tasks. A container instance can be considered as a virtual machine program deployed on a specific server, used to utilize the server's hardware computing resources (such as CPU and memory) to process batch tasks. Generally, a container instance deployed on a server can also be considered as a virtual server obtained by virtualizing that server. Therefore, in this embodiment, a node in the distributed batch system can also be considered as a server included in the distributed batch system for processing batch tasks.

[0052] A topology diagram is a graphical interface used in commercial bank scheduling platforms to display information related to batch jobs. It presents the dependencies and sequential relationships between multiple batch nodes in a tree structure, and can reflect the running status of each node in real time (such as initialization, running, running successfully / failed, skipped, etc.). It provides intuitive monitoring, query and operation support for operators and maintenance personnel, making it easier to manage batch jobs efficiently.

[0053] Weight is a quantifiable measure that is artificially assigned. It reflects the relative importance, reliability, or influence of an object in a system through numerical differences, and ultimately serves data correction, decision optimization, or model learning.

[0054] The comprehensive weight is a core indicator for banks to comprehensively and quantitatively evaluate the importance of nodes in batch systems. The comprehensive weight is generally constructed by integrating technical attributes and business attributes. It is used to identify the comprehensive impact of nodes on system stability in batch links and to provide a quantitative basis for critical path identification and compression decisions.

[0055] Graph compression, also known as path folding or path simplification, is a core concept in many fields of computer science. Its goal is to reduce the number of points that make up a path while preserving the original shape and key features as much as possible. This technique is widely used in computer graphics, geographic information systems (GIS), robot path planning, and data visualization to reduce the complexity of data storage and processing and improve algorithm execution efficiency.

[0056] The core definition of path folding can be understood from the following two levels:

[0057] First, from a graph theory perspective, this involves "folding" a series of intermediate nodes and edges connecting the starting and ending points into direct edges or relationships using specific rules. For example, the path A→B→C can be folded into a single direct edge A→C. This operation is often used to simplify network topology, reduce graph complexity, or provide simpler input for subsequent graph analysis algorithms.

[0058] Second, from a geometric perspective, in geometric applications, paths are typically represented by a series of ordered coordinate points (polylines or polylines). Path folding removes intermediate points that contribute little to the overall shape of the path, retaining only those "key points" that are essential to maintaining the overall shape of the path.

[0059] At any level, the core objective of path folding is:

[0060] First, it significantly reduces the amount of data required to describe the path, achieving data compression;

[0061] Second, reduce the computational burden of subsequent processing (such as rendering, collision detection, and path tracking) to improve efficiency;

[0062] Third, while simplifying, retain the geometric or topological features of the original path as much as possible to avoid excessive distortion and achieve feature preservation.

[0063] Among numerous path folding algorithms, the Douglas-Peucker algorithm (also known as the Ramer-Douglas-Peucker algorithm or the DP algorithm) is highly efficient and widely applicable, and is therefore considered one of the most classic technical solutions. Its implementation principle is detailed below.

[0064] First, the core idea of ​​the DP algorithm.

[0065] The dynamic programming (DP) algorithm employs a strategy of "recursive divide and conquer" and "iterative endpoint fitting." Its core idea is:

[0066] (i) For a curve represented by a sequence of points, first, form a straight line segment by connecting its first and last points (let's call the first and last points P1 and Pn). Then, find the point Pm that is the longest distance from all points on the curve to the straight line segment, and record the distance from Pm to the straight line segment as the longest distance.

[0067] (ii) If the maximum distance is less than the preset threshold ε, then the straight line segment is considered to be a sufficiently good approximation of the original curve, and all intermediate points on the path can be "folded" away, leaving only the first and last endpoints.

[0068] (iii) If this maximum distance is greater than the threshold ε, it indicates that the approximation effect of the straight line segment is not very good, and point Pm is an important "feature point" and should be retained. At this time, the original curve is divided into two sub-curves with Pm as the boundary: the segment from P1 to Pm, denoted as sub-curve [P1, Pm], and the segment from Pm to Pn, denoted as sub-curve [Pm, Pn]. The same processing procedure is recursively applied to these two sub-curves.

[0069] This process continues until all sub-curves satisfy the simplification conditions. Finally, all the retained points (the initial two endpoints and the feature points retained during the recursion) are connected in sequence to form the simplified path.

[0070] Second, the mathematical description of the algorithm.

[0071] The DP algorithm is described below using mathematical language:

[0072] (I) Input and Output:

[0073] Input includes:

[0074] (1) The ordered set of points representing the original path is denoted as C = (P1, P2, ..., Pn), where Pi is a coordinate point denoted as (xi, yi), and i is an integer from 1 to n;

[0075] (2) Distance threshold (also known as error tolerance) ε is generally a real number greater than 0.

[0076] The output includes:

[0077] The simplified ordered node set C' is denoted as C' = (P1, Pk1, ..., Pk). m , Pn), and C' is a subset of C.

[0078] (II) The algorithm steps are as follows:

[0079] Step 1, Initial connection: Connect the first point P1 and the last point Pn of the path with a straight line segment P1-Pn;

[0080] Step 2: Find the farthest point. For each intermediate point between P1 and Pn, calculate its perpendicular distance to the line segment P1-Pn, and find the point Pm with the largest perpendicular distance and its corresponding maximum distance dm.

[0081] Step 3, Threshold Judgment and Recursion: If dm is less than or equal to the distance threshold, it means that the deviation of all intermediate points of the current path from the approximate straight line is within the tolerance range. Therefore, all intermediate points are discarded. The simplified result of this path segment is line segment P1-Pn; if dm is greater than the distance threshold, it means that Pm is a non-negligible "feature point" and must be retained. Using Pm as the dividing point, the original path is divided into two sub-paths, denoted as sub-path C1=[P1, Pm] and sub-path C2=[Pm, Pn]. The former includes all points from P1 to Pm in C, including P1 and Pm, and the latter includes all points from Pm to Pn in C, including Pm and Pn. Then, the recursive algorithm is called on C1 and C2 respectively, that is, steps 1 to 3 are executed recursively.

[0082] Step 4: merging results, that is, merging recursive results to generate a simplified node set.

[0083] The shortcomings of the above-mentioned existing technologies are as follows:

[0084] Firstly, there is a risk of missing key information, as existing path folding solutions rely solely on geometric features for compression. In the operational scenarios of the financial industry, high-value nodes may be treated the same as low-value nodes, posing a risk of mistakenly folding critical business nodes and thus compromising the integrity of core business processes.

[0085] Secondly, static thresholds cannot adapt to heterogeneous topologies. Traditional path folding schemes use fixed lengths and weighted thresholds, which cannot adapt to the heterogeneous topologies of different business lines. In sparse topologies, insufficient compression is likely to occur, resulting in redundant information; in dense topologies, over-compression is likely to occur, causing loss of effective information, thereby significantly reducing the efficiency of automated processing.

[0086] Thirdly, insufficient information capacity restricts operational efficiency. Existing solutions only retain the geometric relationships of the endpoints of the folded path and cannot inherit the business attributes of intermediate nodes, resulting in a lack of data support for operational decisions. At the same time, the irreversible compressor manufacturing causes the loss of original path information, requiring additional storage to maintain complete topology data, which increases system complexity and makes it difficult to meet the rigid requirements of the financial industry for operational traceability.

[0087] To address the limitations of existing technologies, this application proposes a business value-driven graph compression method (also known as a path folding method) based on pre-calculated weights. With the comprehensive weights of nodes and the critical path already determined, a dual-threshold mechanism is used to filter low-value long paths. A dynamic adjustment algorithm is combined to adapt to the topology of different business systems, while an aggregated edge metadata storage scheme supports the traceability of the operational process. This solution attempts to effectively simplify the topology while maintaining the integrity of core business processes.

[0088] This application aims to address the low operational efficiency of commercial banks' batch systems due to their complex topologies and large node scale. Traditional graph compression algorithms struggle to adapt to the unique characteristics of banking systems, such as unweighted edges and weighted nodes, and cannot differentiate between business value densities, resulting in poor topology visualization and difficulties in fault location. To address this, this application proposes an adaptive compression method that integrates business semantics. By dynamically folding low-value paths while preserving the integrity of critical business processes, the topology is simplified, improving operational response speed and decision reliability.

[0089] This application first identifies critical paths based on the comprehensive weight of nodes, fully preserving all nodes and connections, and marking the minimum hop count between critical nodes to evaluate path efficiency. For non-critical paths, a dual-threshold screening mechanism is designed: only when the number of intermediate nodes in a path exceeds a preset length threshold and the average node weight is lower than a preset weight threshold is it determined to be a low-value long path and a folding operation is performed. During the folding process, all intermediate nodes and associated edges are removed, and an aggregate edge is added to connect the first and last nodes. The average weight of the folded path is assigned to this aggregate edge to inherit business attributes. To improve cross-business system adaptability, a dynamic threshold optimization algorithm is adopted: based on the user-defined desired compression rate, the threshold parameters are adjusted through iterative feedback to ensure reasonable compression for different topologies. Simultaneously, the unique identifier and average weight of the folded path are stored through the aggregate edge, supporting one-click backtracking of the original topology, meeting the rigid requirements of operational traceability in the financial industry.

[0090] In commercial banks' batch processing systems, the continuity and stability of business processes are crucial. These processes are carried by "critical paths," which consist of a series of nodes with high technical risk and high business value. At the same time, there are a large number of "non-critical paths" in the system, consisting of low-risk, low-business-value nodes. These paths often present a lengthy chain structure. Although they are part of the system topology, in scenarios of macro-monitoring and rapid fault location, their detailed, hop-by-hop topology is not only not necessary information, but also becomes "noise" for visual and analytical purposes, seriously reducing visualization effectiveness and operational efficiency.

[0091] This invention addresses this pain point by focusing graph compression on these "low-value long paths" and performing efficient and reversible compression processing, namely, "path folding." Through path folding, the topology graph is significantly simplified without compromising the integrity of core business processes, thereby improving its loading speed and visualization readability.

[0092] Please see Figure 1 The graph compression method in this embodiment can be divided into four key steps.

[0093] Step one: Identify and retain the critical path. Specifically, all critical paths can be identified based on pre-calculated node weights that integrate technical risks and business value. All nodes and edges on the critical path are fully retained without any compression or aggregation. This ensures that operations personnel can still clearly obtain the full picture of the core business when viewing the compressed topology diagram.

[0094] Step two involves identifying and filtering non-critical paths. Specifically, after retaining the critical paths, the remaining non-critical paths (i.e., paths other than the critical paths) are filtered under two conditions: only low-weight long paths that meet both the length and weight conditions are selected from the non-critical paths. This mechanism ensures that only those paths with "low value density" and "high visual redundancy" are collapsed, avoiding the miscompression of important local structures that are not on the critical paths but have high weight or short length.

[0095] The length condition refers to the number of nodes on the path exceeding a preset length threshold Lth; the weight condition refers to the average comprehensive weight of the nodes on the path being lower than a preset weight threshold Wth.

[0096] Step 3: Perform path folding. Specifically, for the selected low-weight long paths, path folding will be performed. Specifically, assume a low-weight long path Q to be folded consists of a first node u1 and a last node u2. m And the m-2 consecutive nodes in the middle (u2, u3...u...) m-1 The graph is composed of nodes, and collapsing this low-weight long path removes all intermediate nodes and the edges connecting them from the graph. In their place are the first node u1, the last node u2, and the last node u3. m And a newly created shortcut edge that directly connects the first and last nodes.

[0097] Step four: Calculate the weights after folding. Specifically, considering the characteristic of commercial bank batch systems where edges have no weights but nodes have weights, this application designs an aggregated edge weight inheritance mechanism. This mechanism inherits the weights of all intermediate nodes to be folded, such as the m-2 consecutive nodes (u2, u3...u ... m-1 The average of the comprehensive weights of the newly generated shortcut edges is assigned to them, allowing maintenance personnel to quickly determine the approximate importance of the collapsed parts by viewing the weights of the shortcut edges, and providing a data foundation for subsequent analysis and visualization rendering.

[0098] Before explaining the specific implementation process, for convenience, let's first introduce the basic definitions and symbols that may be used in the specific implementation process.

[0099] First, let the original operation and maintenance topology be G={V, E}, where V is the set of all batch nodes (also called nodes) in the batch system, denoted as the node set, and E is the set of all edges in the original operation and maintenance topology, denoted as the edge set. Edges represent the data connection relationship between nodes, and the direction of the edge indicates the direction of data flow between nodes. For example, if the data output by node A is sent to node B for further processing, then there can be an edge connecting node A and node B in the original operation and maintenance topology, and this edge points from node A to node B.

[0100] Secondly, for any node v in V, define its comprehensive weight w. v The value is greater than 0. The specific value can be set by the bank's operations and maintenance personnel based on the technical and business attributes of the node, without any limitation.

[0101] Third, define the critical path P. crit = (p1, p2, ... p) n ), where p1 to p n The critical path in the original operations and maintenance topology diagram contains n nodes, where n is the length of the critical path. This value varies depending on the critical path and is not limited. The original operations and maintenance topology diagram can have one or more critical paths, with no limit on the specific number. The critical path can be manually set by operations and maintenance personnel or determined using methods for identifying critical paths in relevant technologies; details will not be elaborated further.

[0102] Fourth, define the non-critical path Q = (u1, u2, u3...u... m-1 u m ), where u1 to u m For the m consecutive nodes contained in the non-critical path in the original operation and maintenance topology diagram, u2 to u m-1Let m-2 be the m-2 consecutive intermediate nodes of the non-critical path Q. m-2 represents the length of the non-critical path, and the specific value of m can vary depending on the non-critical path and is not limited. A non-critical path refers to any path in the original operational topology graph that is not the critical path. In the original operational topology graph, a path is formed by all nodes connected sequentially by directed edges.

[0103] Specifically, the original operation and maintenance topology graph is a directed acyclic graph. In a directed acyclic graph, for any sequence of n nodes (u1, u2, ... u... n If the node sequence satisfies the following condition, then the node sequence and the edge connecting every two adjacent nodes in the node sequence constitute a path in this directed acyclic graph:

[0104] The first condition is that for the first node (first node) u1 in the node sequence, there is no edge pointing to the first node, that is, there are no other nodes pointing to the first node u1;

[0105] The second condition is that for the last node (tail node) u in the node sequence n There are no edges originating from this node pointing to other nodes; that is, the tail node does not point to any other node.

[0106] The third condition is that for any two adjacent nodes u in the node sequence... i and u i+1 In a directed acyclic graph, there exists an edge pointing from one element to another.

[0107] The following describes the specific implementation process of the graph compression method provided in the embodiments of this application.

[0108] The first step is to preserve the critical path and mark the minimum number of hops between every two critical nodes in the critical path.

[0109] Preserving all nodes and edges on the critical path helps ensure the integrity of core business processes. Therefore, this application does not compress or merge any nodes and edges on the critical path, but only marks the minimum hop count between critical nodes. The minimum hop count refers to the shortest path length from one critical node to the next in the critical path. The core objective of the first step is to preserve the complete structure of the critical path, facilitating rapid problem identification by operations personnel. Simultaneously, marking the minimum hop count helps assess path efficiency and avoids performance degradation due to path redundancy.

[0110] Critical nodes are the nodes contained in the critical path. The critical path P, as defined above... crit As an example, any two key nodes p i and p j Minimum number of hops between MinHops (pi p j ) can be calculated using the following formula (1), where i and j are both integers between 1 and n, and i and j are not equal, i being less than j.

[0111]

[0112] Where Pkey represents the set of all critical paths in the original operation and maintenance topology diagram.

[0113] The second step is to identify whether each non-critical path meets the length and weight conditions in order to determine the low-weight long path among all non-critical paths.

[0114] Low-weight long paths are non-critical paths that satisfy both the length and weight conditions.

[0115] For a non-critical path, if the length Len(Qmid) of the non-critical path Qmid satisfies the inequality shown in formula (2), then the non-critical path satisfies the length condition.

[0116]

[0117] The length of a non-critical path, Len (Qmid), can be defined as the number of intermediate nodes contained in this non-critical path. For example, in the previous example, the length of the non-critical path, Qmid, is equal to m-2. Lth is a preset minimum path length threshold (also referred to as the length threshold), and its specific value can be calculated using an adaptive algorithm for dynamic thresholds.

[0118] For a non-critical path, if the average weight value Wavg(Qmid) of the non-critical path Qmid satisfies the inequality shown in formula (3), then the non-critical path is determined to satisfy the weight condition.

[0119]

[0120] Wherein, the average weight value Wavg(Qmid) is defined as the average of the combined weights of all intermediate nodes on the non-critical path, w(u i ) represents the i-th node u on the non-critical path. i The comprehensive weight is Wth, which is the preset maximum average weight threshold (or simply weight threshold). Its specific value can be calculated using an adaptive algorithm for dynamic thresholds.

[0121] The third step is to perform path folding on low-weight long paths.

[0122] Assume the non-critical path Q = (u1, u2, u3, ..., u4) m-1 u mQ is identified as a low-weight long path. After path folding is performed on Q, the original operation and maintenance topology will be updated into a new topology. The new topology will have the following two changes compared to the original operation and maintenance topology.

[0123] The first aspect is the change in the node set, which involves filtering all intermediate nodes of the long path with low weight from the node set V of the original operation and maintenance topology.

[0124] The second aspect involves changes to the edge set. From the original edge set E of the operation and maintenance topology graph, all edges constituting the low-weight long path Q are selected, and a new shortcut edge Es is added. This shortcut edge directly connects the first and last nodes of the low-weight long path Q, for example, directly connecting the aforementioned u1 and u... m And it points from the first node to the last node.

[0125] The shortcut edge Es is assigned a shortcut edge weight w(Es) to represent the information of the collapsed low-weight long path Q. w(Es) can be calculated according to the following formula (4).

[0126]

[0127] In practical applications, batch systems across different business lines exhibit significant differences in their topology and node weight distribution. While fixed length thresholds Lth and Wth may be applicable in a specific batch system, they can easily lead to insufficient or excessive compression in heterogeneous topologies. Therefore, this application designs a dynamic adaptive threshold algorithm to improve universality and robustness.

[0128] The necessity and objective of the adaptive algorithm for dynamic thresholds are explained below.

[0129] In terms of necessity, if a static fixed threshold is used, there is a defect in adaptability to heterogeneous topologies. Specifically, sparse topologies with long chains and dense topologies with short paths require different compression strategies. Therefore, the traditional static fixed threshold cannot be universally applied to topologies with different topological structures. Thus, it is necessary to use an adaptive algorithm with a dynamic threshold to determine the dynamic threshold.

[0130] The goal of the adaptive algorithm for dynamic thresholds is to adaptively solve for the optimal threshold pair based on the current input topology graph G and the user-defined target compression ratio (TCR), that is, to solve for the optimal length threshold and weight threshold, so that the compression result satisfies the TCR while maximizing the preservation of the integrity of the original operation and maintenance topology graph compressed in this embodiment.

[0131] See Figure 2The implementation principle of the adaptive algorithm for dynamic thresholds in this embodiment lies in establishing a feedback control loop. First, the threshold is initialized based on global statistics of the graph. Then, compression is performed and the actual compression ratio is calculated. The actual compression ratio is compared with the expected compression ratio, and the threshold is iteratively adjusted according to the error until it converges to the expected target.

[0132] First, let's explain the symbols and definitions that may be involved in the adaptive algorithm for dynamic thresholds.

[0133] The following are the global statistical characteristics of the topology graph.

[0134] Average weight U of non-critical nodes w This represents the average weight of all non-critical nodes in the latest topology graph after path folding. Non-critical nodes refer to nodes belonging to non-critical paths.

[0135] Sigma of weights for non-critical nodes w , which represents the standard deviation of the combined weights of all non-critical nodes in the latest topology graph after path folding.

[0136] Average length U of non-critical paths L , which represents the average path length Len(Qmid) of all non-critical paths in the batch system.

[0137] Total number of non-critical nodes N nocrit , representing the total number of all non-critical nodes in the original operation and maintenance topology diagram, i.e., the original operation and maintenance topology diagram has a total of N. nocrit Non-critical nodes.

[0138] Total number of collapsed nodes N FOLD , representing the total number of nodes that have been collapsed in the latest topology graph (also known as the compressed topology graph) compared to the original operation and maintenance topology graph after path folding. For example, if 20 nodes are removed from the original operation and maintenance topology graph to obtain the latest topology graph through path folding, then the total number of collapsed nodes N is... FOLD It equals 20.

[0139] The actual compression ratio ACK represents the total number of folded nodes N. FOLD The proportion relative to the total number of non-critical nodes can be specifically equal to the ratio obtained by dividing the former by the latter, i.e., ACK=N FOLD ÷N nocrit .

[0140] Error rate is defined as the difference between the expected compression rate (TCR) and the actual compression rate. The expected compression rate is a fixed value set by the maintenance personnel as needed, i.e., Error = TCR - ACK.

[0141] The following are the fixed parameters set by the user (i.e., the operations and maintenance personnel).

[0142] The expected compression ratio (TCR) represents the percentage of non-critical nodes that the user expects to reduce through path folding. For example, a TCR of 30% means that the user expects to remove 30% of the non-critical nodes from the original operation and maintenance topology through path folding.

[0143] The relationship between path folding and graph compression is that graph compression is achieved by folding one or more non-critical paths in the topology graph. In other words, path folding and graph compression can be considered as two equivalent concepts.

[0144] The following are the internal parameters of the adaptive algorithm for dynamic thresholding.

[0145] Length adjustment step A L and weight adjustment step size A W The former controls the adjustment range of the length threshold in each iteration, while the latter controls the adjustment range of the weight threshold in each iteration.

[0146] The following are the inputs and outputs of the adaptive algorithm for dynamic thresholds.

[0147] The input includes a directed acyclic graph G and the desired compression ratio TCR, such as the original operation and maintenance topology graph or the compressed topology graph mentioned above;

[0148] The output includes optimal threshold pairs, which include length thresholds and weight thresholds.

[0149] The following is combined Figure 2 Explain how the adaptive algorithm for dynamic thresholds is implemented.

[0150] Calculate the average weight U of all non-critical nodes in a directed acyclic graph. w and weighted standard deviation Sigma w .

[0151] Calculate the average length U of all non-critical paths in a directed acyclic graph. L .

[0152] Set initial length threshold Lth0 and initial weight threshold Wth0. The initial length threshold can be equal to the average length of non-critical paths rounded down, and the initial weight threshold can be equal to the average weight minus 0.5 times the standard deviation of the weights, which is equal to U. w -0.5Sigma w Setting the initial weight threshold in this way can tend to compress paths with lower weights first.

[0153] Set an iteration counter k and initialize the value of k to 0, i.e., k=0. Set the maximum number of iterations Kmax for the iteration counter. The specific value can be set as needed, for example, Kmax can be set to 10, 20, etc.

[0154] If the preset iteration convergence condition is not met, and the current value of the iteration counter k is less than Kmax, then the following iteration process is executed:

[0155] (1) Perform compression, use the current length threshold and the current weight threshold to filter low-weight long paths from the original operation and maintenance topology map, and perform path folding on the filtered low-weight long paths to obtain the compressed topology map.

[0156] (2) Calculate the error rate of the compressed topology graph.

[0157] (3) Check whether the iterative convergence condition has been met based on the error rate.

[0158] Specifically, if the absolute value of the error rate is less than the preset convergence threshold, it can be determined that the iterative convergence condition has been met, and the current length threshold and the current weight threshold are the optimal solutions. At this time, step (4) can be executed.

[0159] If the absolute value of the error rate is greater than or equal to the preset convergence threshold, it can be determined that the iterative convergence condition has not been met, and the current length threshold and the current weight threshold need to be adjusted. Specifically, step (5) can be executed to make the adjustment.

[0160] The convergence threshold is a fixed value preset by the operations and maintenance personnel, which can generally be set to 0.01.

[0161] (4) Output the current length threshold and the current weight threshold as the aforementioned optimal threshold pair, and the adaptive algorithm for dynamic threshold ends.

[0162] (5) Adjust the current weight threshold and the current length threshold.

[0163] If the absolute value of the error rate is greater than or equal to the convergence threshold, and the error rate is greater than 0, it indicates that the current threshold is too strict, causing many paths to be "not compressed when they should be." In this case, it is necessary to increase the weight threshold so that the average weight of more paths meets the condition, and decrease the length threshold so that more shorter paths can also participate in compression.

[0164] If the absolute value of the error rate is greater than or equal to the convergence threshold, and the error rate is less than 0, it means that the compression is excessive. The threshold is too loose, causing many paths to be compressed even when they should not be. In this case, it is necessary to reduce the weight threshold to reduce the number of compressed paths and increase the length threshold to exclude some paths that are not long enough.

[0165] Therefore, in step (5), the current length threshold can be adjusted according to the following formula (5).

[0166]

[0167] Among them, Lth k+1 This represents the adjusted length threshold, where k represents the k-th iteration, and Error k This represents the error rate calculated during the k-th iteration. [Lth] k (1-A) L Error k The parentheses indicate that the value within the parentheses is rounded down. The max() function takes the maximum value among the multiple values ​​within the parentheses. Formula (5) ensures that the adjusted length threshold is not less than 1 and is an integer.

[0168] The current weight threshold can be adjusted according to the following formula (6).

[0169]

[0170] (6) Increment the iteration counter by 1 and return to step (1) until the preset iteration convergence condition is met, or the current value of the iteration counter k is greater than or equal to Kmax, and then end the iteration.

[0171] The current length threshold and current weight threshold used in the first iteration are the same as the initial length threshold and initial weight threshold mentioned above. The current length threshold and current weight threshold used in the second and subsequent iterations are the adjusted length threshold and adjusted weight threshold from the previous iteration.

[0172] If the iteration ends because the current value of the iteration counter k is greater than or equal to Kmax, then the adjusted length threshold and the adjusted initial threshold obtained in the last iteration are the optimal threshold pair output by the adaptive algorithm of the dynamic threshold in this embodiment.

[0173] In summary, the graph compression method of this embodiment has at least the following advantages.

[0174] Firstly, critical business processes are fully preserved. By retaining all high-value critical paths and using a dual-threshold filtering mechanism to precisely compress low-value long paths, the complexity of the topology diagram is significantly reduced. While eliminating visual interference, this ensures that operations and maintenance personnel can quickly locate core fault points and shorten system anomaly response time.

[0175] Secondly, the dynamic adaptive heterogeneous system dynamically adjusts the path length threshold and weight threshold based on the user-defined desired compression rate using an iterative feedback algorithm. This mechanism enables the compression intensity to automatically adapt to the topological characteristics of different business lines, maintaining information integrity while achieving the target compression rate.

[0176] Third, quantifiable data support is built by aggregating the average weight value of collapsed paths through edge inheritance, transforming massive amounts of node information into visualized metrics. Operations personnel can quickly assess the business impact of compressed areas based on the aggregated edge weights, providing data support for system optimization and avoiding decision-making delays caused by information overload in traditional solutions.

[0177] The key to the graph compression method in this embodiment lies in:

[0178] First, a path folding mechanism based on business value perception. While fully preserving the critical path and marking the minimum hop count between nodes, low-value long paths are selected for folding operations through dual constraints of path length thresholds and node weight thresholds. This mechanism incorporates pre-calculated comprehensive node weights into the compression decision, preventing high-risk nodes from being incorrectly compressed. During the folding process, intermediate nodes and associated edges are removed, and new aggregate edges are added to connect the first and last nodes, storing the unique identifier of the path in the aggregate edge.

[0179] Second, a dynamic threshold iterative algorithm. A feedback control system is constructed based on the desired compression ratio (TCR): the path length threshold and node weight threshold are initialized using global statistical features such as the average weight of non-critical nodes and the average path length. The thresholds are then dynamically adjusted based on the error between multiple compression results and the TCR until convergence. This algorithm overcomes the limitations of static thresholds in adapting to heterogeneous topologies, providing adaptive compression intensity adjustment for different financial business systems.

[0180] Third, the architecture for inheriting business attributes of aggregated edges. Addressing the unweighted nature of edges in banking systems, the average comprehensive weight of intermediate nodes along the path is assigned to newly generated aggregated edges, allowing the simplified topology to retain the ability to assess the business impact of compressed areas. The unique identifier stored in the aggregated edges supports the original path backtracking function, meeting the compliance requirements of financial businesses for operational traceability.

[0181] Based on the above methods, this application provides an image compression method. Please refer to [link to relevant documentation]. Figure 3 The method may include the following steps.

[0182] S301, determine the critical and non-critical paths of the original operation and maintenance topology graph. The original operation and maintenance topology graph includes multiple nodes of the distributed system and edges representing the connection relationships between the nodes.

[0183] Non-critical paths refer to paths other than the critical paths in the original operation and maintenance topology diagram. Methods for determining critical paths include summing the comprehensive weights of all nodes in each path to obtain the cumulative comprehensive weight value for each path, and identifying paths with a cumulative comprehensive weight value greater than a preset critical path threshold as critical paths; or sorting all paths in descending order according to their cumulative comprehensive weight values ​​and selecting the top C paths as critical paths, where C is a preset integer greater than 1, and the specific value is not limited.

[0184] Another method for determining the critical path is to calculate the average of the comprehensive weights of all nodes in the path for each path, obtain the average comprehensive weight of each path, and determine the path whose average comprehensive weight is greater than the preset critical path threshold as the critical path. Alternatively, after sorting all paths in descending order according to the average comprehensive weight, select the first C paths as the critical paths, where C is a preset integer greater than 1, and the specific value is not limited.

[0185] In other alternative embodiments, the comprehensive weight of the nodes contained in each path can be analyzed based on other methods to determine the critical path, and no specific method is limited.

[0186] S302, while maintaining the critical path, performs path folding on the non-critical paths to obtain a compressed topology graph.

[0187] S303, in response to the display operation command, displays the compressed topology map.

[0188] Display operation commands can be any operation command determined based on the user's operation to trigger the display of the compressed topology map. For example, display operation commands can be operation commands where the user presses a specified physical button on the electronic device, operation commands where the user clicks a virtual control on the display screen, or other forms of operation commands, with no specific limitation on the form.

[0189] The compressed topology diagram can be displayed directly on the display screen of the electronic device executing the graph compression method of this embodiment, or it can be displayed on any personal terminal device used by maintenance personnel, and the specific display method is not limited.

[0190] The beneficial effects of this solution are as follows: Before compressing the original operation and maintenance topology map, critical and non-critical paths are pre-determined, and path folding is performed only on non-critical paths while the critical paths remain unchanged. This not only compresses the original operation and maintenance topology map through path folding, reducing its complexity, but also avoids the loss of important information caused by folding critical paths. Thus, the problem of information loss is avoided while making it convenient for users to view the topology map.

[0191] Optionally, in step S302, non-critical paths are folded to obtain a compressed topology map, including:

[0192] In non-critical paths, filter out low-weight long paths whose length is greater than the length threshold and whose average weight value is less than the weight threshold.

[0193] Path folding is performed on low-weight long paths to obtain a compressed topology graph.

[0194] The method for filtering low-weight long paths can be found in the second step of the aforementioned embodiment, which identifies whether each non-critical path meets the length and weight conditions in order to determine the low-weight long paths among all non-critical paths.

[0195] Optionally, path folding is performed on low-weight long paths to obtain a compressed topology graph, including:

[0196] Remove all intermediate nodes from the node set of the original operation and maintenance topology graph. Intermediate nodes include all nodes in the low-weight long path except for the first and last nodes.

[0197] From the edge set of the original operation and maintenance topology graph, remove all edges contained in the low-weight long path, and add shortcut edges that directly connect the first and last nodes of the low-weight long path to obtain the updated edge set; the shortcut edge has a shortcut edge weight, which is determined based on the comprehensive weight of the intermediate nodes of the low-weight long path.

[0198] The compressed topology graph is formed by the node set after deleting all intermediate nodes of the low-weight long path and the updated edge set.

[0199] For a detailed method of performing path folding on low-weight long paths, please refer to step three of the aforementioned embodiments, which describes performing path folding on low-weight long paths.

[0200] Optionally, before performing path folding on non-critical paths to obtain the compressed topology map, the following steps are also included:

[0201] To achieve the preset desired compression ratio;

[0202] The initial length threshold and initial weight threshold are iteratively updated based on the desired compression ratio to obtain the optimal threshold pair, which includes the length threshold and weight threshold used to filter out long paths with low weight.

[0203] The method for determining the optimal threshold pair described above can be found in the adaptive algorithm for dynamic thresholds in the foregoing embodiments.

[0204] Optionally, display the compressed topology map, including:

[0205] Display the compressed topology graph and label the minimum hop count for the critical path in the compressed topology graph. The specific method for determining the minimum hop count can be found in formula (1).

[0206] This application also provides an image compression device; please refer to [link to relevant documentation]. Figure 4 ,include:

[0207] The determining unit 401 is used to determine the critical path and non-critical path of the original operation and maintenance topology graph, which includes multiple nodes of the distributed system and edges representing the connection relationship between the nodes.

[0208] The folding unit 402 is used to fold non-critical paths while maintaining the critical path to obtain a compressed topology map.

[0209] Display unit 403 is used to display the compressed topology map in response to a display operation command.

[0210] Optionally, the folding unit 402 performs path folding processing on non-critical paths to obtain a compressed topology map, including:

[0211] In non-critical paths, filter out low-weight long paths whose length is greater than the length threshold and whose average weight value is less than the weight threshold.

[0212] Path folding is performed on low-weight long paths to obtain a compressed topology graph.

[0213] Optionally, folding unit 402 performs path folding processing on low-weight long paths to obtain a compressed topology graph, including:

[0214] Remove all intermediate nodes from the node set of the original operation and maintenance topology graph. Intermediate nodes include all nodes in the low-weight long path except for the first and last nodes.

[0215] From the edge set of the original operation and maintenance topology graph, remove all edges contained in the low-weight long path, and add shortcut edges that directly connect the first and last nodes of the low-weight long path to obtain the updated edge set; the shortcut edge has a shortcut edge weight, which is determined based on the comprehensive weight of the intermediate nodes of the low-weight long path.

[0216] The compressed topology graph is formed by the node set after deleting all intermediate nodes of the low-weight long path and the updated edge set.

[0217] Optionally, before performing path folding processing on non-critical paths to obtain the compressed topology map, the folding unit 402 is also used for:

[0218] To achieve the preset desired compression ratio;

[0219] The initial length threshold and initial weight threshold are iteratively updated based on the desired compression ratio to obtain the optimal threshold pair, which includes the length threshold and weight threshold used to filter out long paths with low weight.

[0220] Optionally, display unit 403 displays the compressed topology map, including:

[0221] Display the compressed topology graph and label the minimum hop count of the critical path in the compressed topology graph.

[0222] The specific working principle of the image compression device in this embodiment can be found in the relevant steps of the image compression method in the foregoing embodiment, and will not be repeated here.

[0223] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0224] For ease of description, the above systems or devices are described separately as various modules or units based on their functions. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware components.

[0225] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0226] Finally, it should be noted that in this document, relational terms such as first, second, third, and fourth are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0227] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A graph compression method, characterized in that, include: Determine the critical and non-critical paths of the original operation and maintenance topology graph, which includes multiple nodes of the distributed system and edges representing the connection relationships between the nodes; While maintaining the critical path, the non-critical paths are folded to obtain a compressed topology map; In response to a display operation command, the compressed topology map is displayed.

2. The method according to claim 1, characterized in that, The step of performing path folding on the non-critical paths to obtain a compressed topology map includes: In the non-critical paths, select low-weight long paths whose length is greater than the length threshold and whose average weight value is less than the weight threshold. The low-weight long paths are folded to obtain a compressed topology graph.

3. The method according to claim 2, characterized in that, The step of performing path folding on the low-weight long paths to obtain a compressed topology graph includes: From the node set of the original operation and maintenance topology, delete all intermediate nodes of the low-weight long path. Intermediate nodes include nodes in the low-weight long path other than the first and last nodes. From the edge set of the original operation and maintenance topology graph, delete all edges contained in the low-weight long path, and add shortcut edges that directly connect the first and last nodes of the low-weight long path to obtain the updated edge set; the shortcut edges have shortcut edge weights, which are determined based on the comprehensive weights of the intermediate nodes of the low-weight long path. The node set after deleting all intermediate nodes of the low-weight long path and the updated edge set constitute the compressed topology graph.

4. The method according to claim 2, characterized in that, Before performing path folding on the non-critical paths to obtain the compressed topology map, the method further includes: To achieve the preset desired compression ratio; The initial length threshold and the initial weight threshold are iteratively updated based on the desired compression ratio to obtain an optimal threshold pair, which includes a length threshold and a weight threshold for filtering low-weight long paths.

5. The method according to claim 1, characterized in that, The display of the compressed topology map includes: Display the compressed topology graph and label the minimum hop count of the critical path in the compressed topology graph.

6. A graph compression device, characterized in that, include: A determining unit is used to determine the critical and non-critical paths of the original operation and maintenance topology graph, which includes multiple nodes of the distributed system and edges representing the connection relationships between the nodes; The folding unit is used to fold the non-critical paths while maintaining the critical paths, so as to obtain a compressed topology map. The display unit is used to display the compressed topology map in response to a display operation command.

7. The apparatus according to claim 6, characterized in that, The folding unit performs path folding processing on the non-critical paths to obtain a compressed topology map, including: In the non-critical paths, select low-weight long paths whose length is greater than the length threshold and whose average weight value is less than the weight threshold. The low-weight long paths are folded to obtain a compressed topology graph.

8. The apparatus according to claim 7, characterized in that, The folding unit performs path folding processing on the low-weight long paths to obtain a compressed topology graph, including: From the node set of the original operation and maintenance topology, delete all intermediate nodes of the low-weight long path. Intermediate nodes include nodes in the low-weight long path other than the first and last nodes. From the edge set of the original operation and maintenance topology graph, delete all edges contained in the low-weight long path, and add shortcut edges that directly connect the first and last nodes of the low-weight long path to obtain the updated edge set; the shortcut edges have shortcut edge weights, which are determined based on the comprehensive weights of the intermediate nodes of the low-weight long path. The node set after deleting all intermediate nodes of the low-weight long path and the updated edge set constitute the compressed topology graph.

9. The apparatus according to claim 7, characterized in that, Before the folding unit performs path folding processing on the non-critical paths to obtain the compressed topology map, it is also used for: To achieve the preset desired compression ratio; The initial length threshold and the initial weight threshold are iteratively updated based on the desired compression ratio to obtain an optimal threshold pair, which includes a length threshold and a weight threshold for filtering low-weight long paths.

10. The apparatus according to claim 6, characterized in that, The display unit displays the compressed topology map, including: Display the compressed topology graph and label the minimum hop count of the critical path in the compressed topology graph.

Images