Intelligent medical network traffic distribution method and system based on improved augmented path maximum flow algorithm

Abstract

The application discloses a smart medical network flow distribution method and system based on an improved augmented path maximum flow algorithm, relates to the technical field of smart medical treatment, and has the technical scheme as follows: a smart medical network is abstracted into a directed graph model with a business priority attribute, and key vertices capable of maintaining key connectivity after removal are dynamically identified, a feasible component subgraph that takes into account structure simplification and business guarantee is constructed, an augmented path is found based on adaptive weights that fuse medical service quality requirements, and dynamic reconstruction is triggered when the path is not feasible, which on the one hand reduces the calculation complexity of large-scale medical network flow distribution, and on the other hand, through priority perception and dynamic reconstruction, ensures that high-priority businesses such as emergency surgery control signals and emergency image transmission always obtain low-latency and high-reliable paths, effectively solving the technical problem that a traditional maximum flow algorithm is difficult to simultaneously take into account business hierarchical guarantee and calculation efficiency in a complex medical scene.

Description

Technical Field

[0001] This invention relates to the field of smart healthcare technology, and more specifically, to a smart healthcare network traffic allocation method and system based on an improved augmented path maximum flow algorithm. Background Technology

[0002] With the rapid development of smart healthcare, high-real-time and high-reliability services such as remote surgery, emergency image transmission, and real-time data interaction in intensive care are increasing, posing unprecedented challenges to network traffic allocation. Maximum flow algorithms, as the core theory of network traffic scheduling, are classically implemented in two main categories: augmenting path algorithms and pre-flow push algorithms. In augmenting path algorithms, to address the problem of network connectivity loss due to unreasonable path selection, traditional methods generally introduce reverse edge mechanisms. These mechanisms provide an opportunity to reverse course by establishing a residual network to correct traffic allocation errors. However, the existence of reverse edges significantly increases the number of edges and topological complexity of the graph, resulting in enormous computational overhead when processing large-scale networks, making it difficult to meet the real-time response requirements of modern medical networks.

[0003] To address the aforementioned issues, existing technologies have proposed improved solutions to optimize computational efficiency. One such solution is a novel augmenting path maximum flow algorithm, which proposes a key vertex feasible component algorithm. This algorithm constructs a simplified subgraph by identifying and removing key vertices that do not affect global connectivity, thereby finding augmenting paths without relying on reverse edges. While this approach reduces algorithm complexity to some extent, its core objective is to solve a general maximum flow problem, and it is not adapted to the specific needs of smart healthcare scenarios. When such general algorithms are directly applied to medical networks, significant limitations exist: First, the algorithm does not consider the priority differences of medical services, failing to distinguish the urgency of emergency and general services, and making it difficult to guarantee the service quality of critical services such as remote surgical control signals. Second, it lacks the ability to perceive dynamic changes in network load and cannot adaptively adjust the topology simplification strategy according to time-varying load fluctuations such as peak outpatient hours and off-peak hours. Third, when simplifying the network graph, excessive pruning may disrupt the connectivity of critical medical services, or the network may become trapped in local optima when locally saturated, failing to fully utilize remaining bandwidth resources.

[0004] Therefore, researching and designing a smart healthcare network traffic allocation method and system based on an improved augmented path maximum flow algorithm that can overcome the above-mentioned defects is an urgent problem to be solved. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a smart healthcare network traffic allocation method and system based on an improved augmented path maximum flow algorithm. Through priority awareness and dynamic reconstruction, it ensures that high-priority services such as emergency surgery control signals and emergency image transmission always obtain low-latency and highly reliable paths, effectively solving the technical challenge of traditional maximum flow algorithms in balancing service hierarchy protection and computational efficiency in complex medical scenarios.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution: Firstly, a method for traffic allocation in smart healthcare networks based on an improved augmented path maximum flow algorithm is provided, including the following steps: The smart healthcare network is abstracted as a directed graph model with capacity and service priority attributes; In the directed graph model, a list of key vertices is dynamically identified and managed, wherein the key vertices in the list are those that meet preset conditions and whose removal does not affect the critical connectivity of the network. Based on the list of key vertices, a feasible component subgraph is constructed from the directed graph model, and the edges in the feasible component subgraph are assigned adaptive weights that incorporate the service quality requirements of medical services. In the feasible component subgraph, augmenting paths are found for medical service flows based on the adaptive weights. After verification, traffic is allocated along the augmenting paths and the network status is updated. In response to the failure to find the augmenting path in the feasible component subgraph, the network is dynamically reconstructed, the feasible component subgraph and the adaptive weights are updated, and the search for the augmenting path is restarted until the global termination condition is met.

[0007] Furthermore, the directed graph model includes a vertex set, a directed edge set, a capacity function, and a priority function; The vertex set includes a source vertex, a sink vertex, and intermediate vertices; The directed edges in the set of directed edges represent network links between vertices; The capacity function assigns a positive capacity value to each directed edge to characterize the real-time available bandwidth of the link; The priority function assigns a priority to each directed edge.

[0008] Furthermore, the dynamic identification and management of the key vertex list includes: Traverse the set of vertices in the directed graph model and apply topology hub determination, edge protection determination, and load and service value mismatch determination to filter candidate vertices; Perform destructive connectivity verification on the candidate vertices: remove the candidate vertices and their associated edges from the memory copy, verify the connectivity from the source vertex to the critical destination by breadth-first search, and add the candidate vertex to the critical vertex list only if the connectivity is maintained; The list of critical vertices is dynamically updated based on periodic and / or event-triggered mechanisms, removing vertices that are ineffective, have low load, or have an increased proportion of high-priority traffic, and adding vertices that meet the criteria.

[0009] Furthermore, the topology hub determination configuration is as follows: if the sum of the degrees of the vertices is greater than a preset topology hub threshold, then it is marked as a candidate vertex; The edge protection determination configuration is as follows: if the in-degree or out-degree of a vertex is 0 and the service priority of its associated edge is lower than a preset protection threshold, then it is marked as the candidate vertex; The load and value mismatch determination configuration is as follows: if the load rate of a vertex is higher than the first threshold and the corresponding high-priority traffic ratio is lower than the second threshold, then it is marked as the candidate vertex.

[0010] Furthermore, the construction of the feasible component subgraph includes: Remove vertices and their associated edges from the list of key vertices in the directed graph model to form an initial simplified subgraph; For each business flow in the set of critical business flows, verify the connectivity from the source to the destination in the initial simplified subgraph; If there are disconnected business flows, select the minimum number of vertices from the list of critical vertices and re-add them to the initial simplified subgraph to restore the connectivity of all critical business flows and obtain the feasible component subgraph.

[0011] Furthermore, assigning adaptive weights based on the integrated medical service quality requirements to the edges in the feasible component subgraph includes: Define a service quality requirement vector for the service flow, which includes the required thresholds for latency, bandwidth, packet loss rate, and latency jitter. For network link maintenance, a quality of service state vector is provided, which includes the estimated values ​​of corresponding latency, available bandwidth, packet loss rate, and latency jitter. The adaptive weight of the edge to the service flow is calculated by multiplying the static basic weight and the dynamic adaptive factor. The static basic weight is negatively correlated with the estimated latency of the link. The dynamic adaptive factor is obtained by multiplying the exponential adjustment terms of each service quality dimension. The exponential adjustment terms are rewarded or penalized based on the degree to which the link state value deviates from the service demand threshold, and the intensity of the reward or penalty is controlled by a temperature parameter that is dynamically adjusted according to the degree of network congestion.

[0012] Furthermore, the step of finding augmenting paths for medical service flows based on the adaptive weights, verifying the allocation of traffic through the augmenting paths found via the trailing edges, and updating the network state includes: Based on the service quality requirement vector of the business flow to be scheduled, set corresponding adaptive weights for the edges of the feasible component subgraph; Run the minimum weighted path algorithm on the feasible component subgraph with weights set to find candidate paths from the source to the sink. Real-time probe data packets are injected into the candidate paths to measure actual latency and packet loss rate. If the business requirements are met, the path is confirmed as the selected path; otherwise, the relevant edge weights are adjusted and a new path is searched. Calculate the maximum allocatable flow for the selected path, smoothly inject the flow through the token bucket controller, and update the remaining capacity of the path edges and the vertex priority flags.

[0013] Furthermore, the method also includes: When a newly arrived high-priority service flow has no path that meets the requirements in the feasible component subgraph and the remaining network bandwidth is below the threshold, the active low-priority service flows are traversed, and the flow that overlaps with the target path and has the lowest priority or consumes the most resources is selected. Send a pre-preemption notification to the source of the selected flow, release the bandwidth resources it occupies after receiving a response, update the feasible component subgraph, and find augmentation paths for the high-priority service flow again.

[0014] Furthermore, the dynamic reconfiguration includes: Scan the feasible component subgraph and remove all useless vertices and their associated edges whose remaining capacity is zero for either incoming or outgoing edges; Check the connectivity of the reconstructed feasible component subgraph. If the critical business flow is disconnected, select the vertex with the highest connectivity or the highest historical priority from the list of critical vertices and add it to the reconstructed feasible component subgraph to restore connectivity. Based on the recovered feasible component subgraph, key vertex identification, feasible component subgraph construction, and adaptive weight calculation are re-executed.

[0015] Secondly, a smart healthcare network traffic allocation system based on an improved augmented path maximum flow algorithm is provided, including: The model building module is used to abstract the smart healthcare network into a directed graph model with capacity and service priority attributes; The vertex recognition module is used to dynamically identify and manage the list of key vertices in the directed graph model, wherein the key vertices in the list of key vertices are those that meet preset conditions and whose removal does not affect the critical connectivity of the network. The subgraph construction module is used to construct feasible component subgraphs from the directed graph model based on the list of key vertices, and to assign adaptive weights to the edges in the feasible component subgraphs that incorporate the service quality requirements of medical business. The path finding module is used to find augmenting paths for medical service flows based on the adaptive weights in the feasible component subgraph, and after verification, allocate traffic along the augmenting path and update the network status. The reconstruction and update module is used to dynamically reconstruct the network in response to the failure to find the augmenting path in the feasible component subgraph, update the feasible component subgraph and the adaptive weights, and return to search for the augmenting path again until the global termination condition is met.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention provides a smart healthcare network traffic allocation method based on an improved augmenting path maximum flow algorithm. By abstracting the smart healthcare network into a directed graph model with service priority attributes, and dynamically identifying key vertices that can maintain critical connectivity after removal, a feasible component subgraph that balances structural simplification and service assurance is constructed. Then, augmenting paths are found based on adaptive weights that integrate medical service quality requirements, and dynamic reconstruction is triggered when a path is infeasible. On the one hand, this reduces the computational complexity of large-scale medical network traffic allocation. On the other hand, through priority awareness and dynamic reconstruction, it ensures that high-priority services such as emergency surgery control signals and emergency image transmission always obtain low-latency and high-reliability paths. This effectively solves the technical problem that traditional maximum flow algorithms cannot simultaneously balance service hierarchy assurance and computational efficiency in complex medical scenarios.

[0017] 2. This invention selects candidate vertices by integrating topology hubs, edge protection, and load and service value mismatch assessments. It also uses destructive connectivity verification to remove vertices that would cause critical services to be disconnected after removal, ensuring that the critical vertex list only includes high-impact but non-essential nodes. At the same time, based on a periodic or event-triggered dynamic update mechanism, it promptly removes faulty, low-load, or high-priority service-processing vertices and adds vertices that meet the mismatch conditions. This can identify non-critical bottleneck nodes with high loads but mainly carrying low-priority traffic, thus avoiding the disruption of emergency service connectivity due to oversimplification, improving network resource utilization, and ensuring the stability of critical medical services.

[0018] 3. This invention forms an initial simplified subgraph by removing key vertices from the original graph and verifies connectivity for key service flows. Key vertices are restored with a minimum number of vertices only when services are disconnected, thus constructing the final feasible component subgraph. While preserving the core transmission capabilities of the network, redundant nodes and edges are reduced to the greatest extent, thereby reducing the complexity of subsequent path search. At the same time, through the minimum vertex set restoration mechanism, it is ensured that all predefined key services such as emergency image transmission and remote consultation still have at least one feasible path after simplification, avoiding the risk of lifeline service interruption due to topology simplification.

[0019] 4. This invention designs a composite weight for edges by multiplying a static basic weight by a dynamic adaptive factor. The static part encourages the selection of low-latency links, while the dynamic part is based on the deviation between the link status and service requirements in various service quality dimensions. It implements rewards and penalties through an exponential adjustment term, and the intensity of the rewards and penalties is dynamically adjusted according to the degree of network congestion. This weight design prioritizes the removal of low-priority traffic that does not meet the needs of emergency services when the network is congested, and promotes balanced resource utilization when the network is idle. In this way, it continuously meets the low latency and low jitter requirements of services such as remote surgery and real-time monitoring in a dynamically changing network environment.

[0020] 5. This invention verifies actual latency and packet loss rate by injecting real-time probe packets before the selected path. Traffic is allocated only when the measured values ​​meet business requirements; otherwise, the weights of relevant edges are dynamically reduced and a new path is found. At the same time, traffic is smoothly injected through a token bucket controller, and the sending rate is dynamically adjusted according to the queue depth of the path nodes. This effectively eliminates the risk that the actual performance of the theoretically optimal path will not meet the standards due to instantaneous network fluctuations, and avoids interference from sudden traffic to jitter-sensitive services such as surgical video streams.

[0021] 6. This invention identifies and pre-preempts low-priority active flows that overlap with the path when a new high-priority service arrives but there is no path to meet its needs and resources are insufficient. After releasing bandwidth resources, it re-paths the high-priority service. This transforms resource preemption from forced interruption to negotiated pre-notification, which ensures that critical services such as emergency image transmission and surgical control can obtain links first when resources are scarce, while reducing unnecessary interruptions of low-priority services. This enables hierarchical emergency allocation of medical network resources in extreme scenarios. Attached Figure Description

[0022] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart from Embodiment 1 of the present invention; Figure 2 This is a system block diagram in Embodiment 2 of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0024] Example 1: A smart healthcare network traffic allocation method based on an improved augmented path maximum flow algorithm, such as... Figure 1 As shown, this is achieved through the following steps.

[0025] Step S1: Abstract the smart healthcare network into a directed graph model with capacity and service priority attributes.

[0026] S101: Construct a directed graph model of the network.

[0027] A vertex set represents a communication entity or data terminal in a network, comprising a source vertex S, a sink vertex T, and intermediate vertices. The source vertex is the starting point of the data flow, typically corresponding to a core data center, cloud storage pool, or central server; the sink vertex is the destination of the data flow, such as an imaging diagnostic workstation, a mobile medical tablet, or a remote consultation terminal. Intermediate vertices are forwarding and processing nodes in the network, including but not limited to local servers in various specialty departments, edge computing nodes, network switches / routers, security gateways, and data interfaces of medical imaging equipment.

[0028] The smart healthcare network is abstracted as a directed graph model. A smart healthcare network can be a hospital campus network or a regional medical cloud platform; there are no restrictions on its implementation.

[0029] This is a set of directed edges, representing directed network links between vertices, defining the data transmission channels and directions. Each edge... Indicates from vertex To the top A directed link, such as fiber optic cable, dedicated 5G link, Wi-Fi 6 wireless channel, etc.

[0030] Let be the capacity function, characterizing the physical or logical transmission capacity of a network link. Let be the capacity function for each directed edge. Assign a positive capacity value The unit is bps, representing the maximum available bandwidth of the link at the current moment. The capacity value can be obtained in real time or near real time through Network Management Protocol (SNMP) polling, NetFlow / sFlow traffic analysis, or the global view of the Software Defined Networking (SDN) controller, thereby ensuring that the model can reflect the actual load capacity of the network and provide a basis for dynamic traffic allocation.

[0031] This is a priority function that characterizes the degree of importance of the associated services for elements in the network. In this embodiment, the priority function is... Associated with each directed edge Above, recorded as Priority is dynamically or statically mapped based on the type of data service that is expected to be carried or is currently being carried on the link. For example: This indicates the emergency level and is used to carry services that are extremely sensitive to latency and jitter, such as real-time vital signs streams, remote surgical control signals, and emergency CT image query requests. , indicating advanced, is used to carry out important but not millisecond-level response services such as routine medical image transmission and electronic medical record retrieval. , indicating the ordinary level, is used to carry background business that can tolerate a certain delay, such as medical management data synchronization, scientific research data backup, and non-real-time log upload.

[0032] S102: Initialize the data structure.

[0033] The above directed graph model This data is transformed into a computer-processable data structure. In this embodiment, an extended one-dimensional structure array is used for storage.

[0034] The basic structure of the data structure is defined as a structure type Node, and it is represented by a one-dimensional array Node[]. For all vertices in the array, the array index is the unique identifier (ID) of the corresponding vertex.

[0035] For example, nodes[i] represents vertices. , containing list List List List and integer fields Among them, the list Store the IDs of all direct predecessor vertices, such that for all vertices satisfying ... vertex Their IDs are stored in this list. Store the IDs of all direct successor vertices, such that for all vertices satisfying ... vertex Their IDs are stored in this list. With list The elements in the database correspond one-to-one, storing the information from each predecessor vertex to the current vertex. The current remaining capacity of the directed edges, such as Corresponding edges Real-time remaining bandwidth List With list The elements in the database correspond one-to-one, storing data from the current vertex. The current remaining capacity of the directed edges to each successor vertex, such as Corresponding edges Real-time remaining bandwidth Integer field This is used to record the current highest priority of the data flow passing through this vertex. It is typically initialized to 0, indicating no flow at the moment. During flow allocation, when a flow with a priority of [priority value missing] is selected... When data flows through this vertex, this field can be updated to... This field provides state information for subsequent decision-making based on vertex priority in algorithms, such as the optimization of critical vertex selection strategies.

[0036] Step S2: In the directed graph model, dynamically identify and manage the list of key vertices. Key vertices are those that meet preset conditions and whose removal does not affect the critical connectivity of the network.

[0037] S201: Multi-dimensional candidate vertex determination.

[0038] From all vertices of G, vertices that may pose a bottleneck to the overall network throughput or play a critical forwarding role are selected as candidates. Since not all highly connected vertices are critical bottlenecks that need to be managed, the real bottlenecks may appear on nodes where the load is mismatched with the service value. Therefore, this invention adopts a comprehensive judgment framework that integrates topology structure, real-time load, and service priority information.

[0039] Traversing the vertex set Each vertex in The following judgment conditions are applied in sequence: Topological hub determination: Calculate vertices degree and , As vertex The set of predecessor vertices, As vertex The subsequent set of vertices, To retrieve the number of elements in the set. If Then These are marked as candidate vertices. The threshold is configurable, for example, with an initial value of twice the average network degree. This condition is used to identify physical hubs such as core switches and data center gateways.

[0040] Edge protection determination: If the vertex satisfies or Vertices with an in-degree or out-degree of 0 require additional checks on their associated business priorities. If the business priority P on their connecting lines is greater than or equal to 2, they are marked as protected edge nodes and excluded from the candidate vertex list to prevent mismanagement of critical medical data sources / sinks. If the business priority P on their connecting lines is less than 2, they will be... Mark them as candidate vertices.

[0041] Load and value mismatch determination: For a vertex Calculate two indicators periodically: Load rate: ,in, At the current moment, the vertex The sum of the remaining capacity of all incoming edges; At the current moment, the vertex The sum of the remaining capacity of all outgoing edges; For the initialization phase or at the start of the previous network steady-state cycle, the vertex Initial total capacity of all incoming edges; For the initialization phase or at the start of the previous network steady-state cycle, the vertex Initial total capacity of all outgoing edges; As vertex The load rate, a value close to 1 indicates high busyness.

[0042] High-priority traffic percentage: ,in, Within a pre-defined statistical time window, the number of vertices that flow through the area... The total flow of all data streams with priority identifier P greater than or equal to 2; Within a pre-defined statistical time window, the number of vertices that flow through the area... Total traffic of all priority data streams; As vertex The proportion of high-priority traffic reflects the urgency of the business being processed by the node. The closer the ratio is to 1, the more urgent and critical the data flowing through the node is; the closer the ratio is to 0, the more ordinary the business is being processed by the node.

[0043] Decision logic: If and If so, then the vertex is determined to be a candidate vertex. The first threshold corresponding to the load rate. This is the second threshold corresponding to the proportion of high-priority traffic, and all values ​​are less than 1. A node with a high load but mainly transmitting low-priority data can clog the network, but has a limited impact on urgent services. Identifying and managing such a node as a critical vertices can effectively clear the network and release potential bandwidth for high-priority services.

[0044] S202: Verification of candidate vertex connectivity.

[0045] To ensure that the basic service capabilities of the medical network are not disrupted after a vertex marked as a candidate is temporarily removed, thus guaranteeing the logical connectivity of critical services, this invention verifies the robustness of the network topology in a memory copy through simulated destructive testing.

[0046] For all candidate vertices selected based on any of the above criteria, a destructive test is performed to ensure network robustness. First, a temporary copy of G is created in memory. .from Simulation of removing candidate vertices And all its associated edges. After performing the above virtual removal, a graph traversal algorithm such as breadth-first search (BFS) is used to traverse this modified copy. In this process, it is necessary to check whether the data channels from the network source to all critical destinations are still open.

[0047] Only when BFS verifies and confirms that the candidate vertex has been removed, and that no data channels from the network source to all critical destinations have been lost, is the candidate vertex officially marked as a critical vertex and its unique identifier added to the critical vertex list K.

[0048] S203: Dynamic update management of the critical vertex list.

[0049] To enable the critical vertex list to adapt to dynamic changes in network conditions and business models and achieve lifecycle management, an intelligent background update module is established. This module maintains the K-list based on a mechanism combining periodicity and event triggering. The background process operates on a basic periodicity... wake.

[0050] In some optional examples, each vertex in the K list is continuously monitored, and it is immediately removed from the list if either of the following conditions is met: a) the node fails; b) its load rate remains below a threshold. c) The proportion of its high-priority traffic increases and exceeds the threshold for three consecutive monitoring periods; This means that the node has been transformed into one primarily handling critical business processes and should no longer be simplified or removed. If any of these conditions are met, it will be removed from K.

[0051] In some optional examples, during each update cycle, vertices not in K are scanned, and the decision conditions in S201 are applied to them. For newly satisfied vertices, connectivity verification in S202 is performed, and those that pass are added to K.

[0052] In some optional examples, the load on the medical network fluctuates regularly, such as during peak outpatient hours in the daytime and during stable periods at night. A fixed update cycle may result in insufficient response during peak hours and redundant computation during stable periods. Therefore, this invention maintains a historical average network load over a sliding time window, assuming the current average load is... The average load for the same period in history was For example, at the same time yesterday, the update cycle adopts a dynamic update cycle. The calculation formula is as follows: ; in, This is a sensitivity coefficient, and its value can be flexibly set as needed. When the network load is higher than historical levels, it is considered to be entering a peak period. Automatically shortening the cycle speeds up the identification and adjustment of critical vertices to respond quickly to traffic changes; when the load is low, the cycle is lengthened to save system resources, which enables the management strategy to have the ability to sense business rhythm.

[0053] Step S3: Based on the current list of key vertices, construct a feasible component subgraph from the directed graph model, and assign adaptive weights to the edges in the feasible component subgraph that incorporate the service quality requirements of the medical business.

[0054] S301: Construct the initial simplified subgraph.

[0055] While preserving the core connectivity of the network, the complex global network graph is simplified in the first step to reduce the complexity of subsequent path calculation algorithms. From Temporarily remove all vertices and their connected edges from the current critical vertex list K to obtain an initial simplified subgraph. This operation is implemented in memory by modifying the graph data structure, such as clearing the relevant vertices. , The list and its corresponding capacity list represent a logical masking rather than a physical deletion.

[0056] S302: Verification and repair of critical business connectivity.

[0057] To ensure that the simplified subgraph does not compromise the transport reachability of any defined critical medical services, a set F of critical service flows is maintained. Each service flow f in this set is defined as a triple. , representing the source point, destination point, and their multidimensional service quality requirement matrix, respectively.

[0058] In obtaining Afterwards, targeting For each business flow f in the process, use a shortest path algorithm (such as Dijkstra's algorithm) to calculate the shortest path path. The above verification exists from arrive The connection path. If the verification passes, it indicates that the service flow does not depend on the removed critical vertex. If the verification fails, it means that a certain service flow f is in If the connection is no longer established, it means that the removed critical vertices contained essential nodes for the business flow. In this case, the fewest critical vertices from K that can restore the connectivity of this business flow will be selected and rejoined. This operation can be achieved by solving a minimum vertex set problem with the goal of restoring connectivity, or by using heuristic rules, such as prioritizing the restoration of high-priority services.

[0059] This step yields the final feasible component subgraph used for computation. . While maintaining a simplified structure, it ensures that all predefined key medical services have at least one logical pathway, thus avoiding business inaccessibility issues caused by oversimplification.

[0060] S303: Define the business service quality requirement vector.

[0061] A multi-dimensional vector is used to characterize the different sensitivities of medical services to network transmission quality, defining a service quality requirement vector for each type of medical service or each data stream. For example, the service quality requirement vector of business flow f. ,in This indicates the maximum end-to-end latency that the business can tolerate. This indicates the minimum guaranteed bandwidth required for stable service transmission. This indicates the maximum packet loss rate that the business can tolerate. This represents the maximum latency jitter that the service can tolerate. Different services have different vector values, such as those for remote surgical control signals. It may require less than 20 milliseconds, while medical record backup services... A few seconds may be allowed.

[0062] S304: Maintenance link service quality state vector.

[0063] For each edge in the network Maintain a quality of service state vector This reflects the real-time or estimated status of the link, where... This represents the current estimated transmission delay for this link. This represents the currently available bandwidth for this link. To estimate the packet loss rate, To estimate jitter.

[0064] In some optional examples, the above status values ​​can be obtained through active probing by sending probe packets, passive monitoring by analyzing NetFlow data, or feedback from the SDN controller.

[0065] S305: Calculate adaptive weights for converged business requirements.

[0066] By deeply integrating static network topology information with dynamic business requirements and link status, the final cost or attractiveness weight of an edge is generated to guide the algorithm to intelligently select the most suitable path for different businesses.

[0067] Design a weight function: edge Weights for carrying service flow f , by static basic weights With dynamic adaptive factor Joint decision: ; ; ; Among them, static basic weights This reflects the inherent transmission characteristics of the link and encourages the selection of high-quality links. In this embodiment, latency is used as a reference. Based on the reciprocal, the choice of low-latency paths is encouraged. To prevent small constants from being divided by zero, other basic metrics, such as available bandwidth, can be selected based on the main optimization objective. As a dimension of service quality, For the edge In the current state, in dimension The measured or estimated value of the service capacity that can be provided; For business flow f in dimension The minimum demand threshold; The temperature parameter, used to control the penalty / reward steepness, is automatically reduced during network congestion. The value makes the exponent term The penalties for links that fail to meet demand are more severe, thus more aggressively directing traffic to paths that do meet demand, ensuring critical business operations, while automatically increasing [their efficiency] when the network is idle. This value makes the weight distribution smoother, allowing for a more balanced distribution of traffic and improving overall resource utilization. Let f be the importance coefficient of the business flow f in dimension q.

[0068] Step S4: In the feasible component subgraph, find augmenting paths for the medical service flow based on adaptive weights, and after verification, allocate traffic along the path and update the network status.

[0069] S401: Candidate path search based on adaptive weights.

[0070] By using the adaptive weights calculated in step S3, which incorporate the service quality requirements of specific businesses, the path finding problem is transformed into finding the path with the lowest cost in the weighted graph. This guides the algorithm to prioritize links with low latency that best meet the specific requirements of the business.

[0071] When a service flow f to be scheduled arrives, its service quality requirement vector is read. and with feasible component subgraphs Based on this, each of its edges The weights are temporarily set to adaptive weights calculated for this business flow. Subsequently, on this dynamically generated weighted graph, the minimum weight path algorithm is run to find a path with the minimum total weight from the business flow source point S to the sink point T. This path is the candidate path. This path is theoretically optimal, but its feasibility needs to be verified by real-time conditions.

[0072] S402: Real-time performance verification and feedback optimization of the path.

[0073] Considering the risk that the theoretically optimal path selected by the algorithm based on historical or estimated information may fail to meet performance standards during actual transmission due to instantaneous fluctuations in network conditions such as sudden traffic surges and link jitter, this invention introduces an online probe verification mechanism to perform health checks on candidate paths before officially allocating service traffic.

[0074] Obtain candidate paths Immediately afterwards, a low-overhead real-time probe data packet is injected into the path. This real-time probe data packet records the timestamp of each hop on the path, and calculates the current measured performance vector of the path in real time based on its round-trip time or end-to-end time and packet loss. .

[0075] Next, Service quality requirement vector of business flow f Perform a dimension-by-dimensional comparison. Only when Satisfies in all key dimensions Only when the threshold requirement is met will it be... Confirmed as the selected path that can ultimately be used for transmission If the verification fails, feedback and optimization will be triggered: based on... and The gap is dynamically reduced. For China The weight parameters of the relevant edges, such as in the weight formula. The penalty term in the factor is temporarily increased, and then the process returns to S401 to recalculate the next candidate path. This mechanism forms a closed loop of weight-guided path finding, real-time verification, and feedback to optimize the weights, ensuring the final reliability of the path selection.

[0076] S403: Smooth traffic delivery and network status updates.

[0077] For medical services that are sensitive to jitter, such as remote surgical video streams, in order to avoid sudden traffic surges that could cause instantaneous congestion and jitter, this invention employs token bucket traffic shaping technology and updates the network model in real time based on the allocation results, thereby achieving bandwidth allocation in a stable and controllable manner.

[0078] ; For the selected path that has passed verification Calculate the traffic it can carry, which is limited by the edge with the smallest capacity on the path, and to ensure that the allocated traffic does not exceed the actual available capacity of any link segment, the traffic it can carry is... The calculation formula is: in, Representing an edge The current remaining capacity.

[0079] To avoid carrying traffic In a one-time burst injection network, a token bucket controller is established at the source point S for this traffic flow. Tokens are generated and stored in the bucket at a constant rate, and a token must be retrieved from the bucket before a data packet is sent. The token generation rate is initially set to a conservative value, and data is smoothly injected into the selected path according to the rate controlled by the token bucket. Simultaneously, monitor the queue depth of key nodes in the path: if queue growth is detected, indicating potential congestion, dynamically reduce the token generation rate; if the queue remains empty, appropriately increase the rate until the cumulative sending volume equals... This is sufficient to mitigate sudden surges in traffic.

[0080] After traffic allocation is completed, the network model is updated to reflect the latest resource usage, providing an accurate basis for the next scheduling. First, for the selected path... Each edge on With the current remaining capacity Subtract the carrying capacity Then, the updated remaining traffic is obtained. Additionally, the selected path... All vertices Field, updated to ;in, It is the priority level of business flow f.

[0081] 404: Iterative loop and local termination judgment.

[0082] After completing a traffic allocation step S401-S403, immediately return to step S401 to attempt to find a new augmenting path for the next scheduled service flow in the updated network state, or continue searching for the next parallel path for the current service flow. When in the current feasible component subgraph... If, during the process, no augmenting path that can be verified in real time can be found for any pending service flow, it is determined that in the current... The traffic allocation under the given conditions reaches local saturation, and the execution ends.

[0083] In some optional examples, if the determination If there is no valid path in the network that can meet the quality of service requirements of the newly arriving service flow, and the remaining available bandwidth of the entire network or a local area is detected to be lower than the set threshold, then the flow pre-preemption mechanism is triggered.

[0084] First, it iterates through currently active service flows in the network that have overlapping paths with the paths required by the new service and have low priority. It selects the lowest priority flow, or, among flows of the same priority, the flow occupying the most overlapping resources. Then, it sends a standardized pre-preemption notification to the source of the selected flow. This pre-preemption notification is a request with a diplomatic protocol, containing information such as the amount of resources to be released and the time window, rather than a forced interruption command. After receiving the notification, the source of the selected flow can choose to suspend data transmission, reduce the transmission rate, or switch to a standby low-bandwidth mode and send an acknowledgment response to the system. Upon acknowledgment, the bandwidth resources released by the source of the selected flow are updated in the remaining capacity of all relevant edges on its occupied path; feasible component subgraph. After the update, augmenting paths are re-found for high-priority business flows.

[0085] Step S5: In response to the inability to find an effective augmenting path in the current feasible component subgraph, the network is dynamically reconstructed, the feasible component subgraph and adaptive weights are updated, and the process returns to step S4 until the global termination condition is met.

[0086] S501: Detection and pruning of useless vertices.

[0087] When step S4 terminates due to the inability to find an augmenting path, immediately scan the current feasible component subgraph. All vertices in the current feasible subgraph. For each vertex, check all its associated directed edges. If the remaining capacity of all incoming edges of a vertex is zero, or the remaining capacity of all outgoing edges of a vertex is zero, then mark the vertex as a useless vertex. Subsequently, from the current feasible subgraph... The vertex and all its connected edges are permanently deleted. This invention removes dead nodes with no data throughput capacity in locally saturated networks, making the graph structure more streamlined and eliminating obstacles for subsequent path discovery.

[0088] S502: Dynamic reconstruction of feasible components.

[0089] Because pruning operations can disrupt graph connectivity, causing critical service flows to lose their transmission paths, it is necessary to select the most suitable vertices from the list of critical vertices and re-add them to restore or enhance the network's transmission capacity in the most economical way.

[0090] like If connectivity is lost, one or more vertices are selected from the list of critical vertices and re-added to the graph. Connectivity is restored with minimal cost, prioritizing critical business processes. For example, the critical vertices with the highest connectivity or the highest historical priority are selected for re-addition. Specifically: After pruning useless vertices, check the current feasible component subgraph. The connectivity. If If connectivity from the source point S to all key business destinations is still maintained, return directly to step S4 and continue searching for augmenting paths in the simplified graph.

[0091] In some optional examples, each candidate vertex in the list of computationally critical vertices with the highest connectivity is compared with the current feasible component subgraph. The vertex with the most potential connecting edges between vertices is selected.

[0092] In some optional examples, the highest historical priority record is the reference vertex. Historical records: Select the vertices that have processed high-priority transactions.

[0093] Re-add the selected key vertices and their associated edges. Subsequently, based on the reconstructed graph, the process jumps back to step S2 and re-executes the dynamic key vertex identification and update process to generate a new list of K values ​​matching the current network state. Then, the process jumps to step S3 to construct a new feasible component subgraph based on the new K values. And calculate adaptive weights.

[0094] S503: Substitute loop and global termination.

[0095] When the list of key vertices K is empty, and in the current feasible component subgraph The algorithm terminates globally when no augmenting path from source S to sink T can be found. At this point, all possible combinations of critical vertices and path assignments have been tried, and the resulting sum of flows is the approximate maximum flow after considering business priorities and constraints under the current network topology and business requirements. This ensures that the algorithm does not loop infinitely, but terminates naturally after exploring all valuable optimization space.

[0096] Example 2: A smart healthcare network traffic allocation system based on an improved augmented path maximum flow algorithm. This system is used to implement the smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm described in Example 1, such as... Figure 2 As shown, it includes a model building module, a vertex recognition module, a subgraph building module, a path finding module, and a reconstruction and update module. The system comprises the following modules: a model building module, which abstracts the smart healthcare network into a directed graph model with capacity and service priority attributes; a vertex identification module, which dynamically identifies and manages a list of key vertices in the directed graph model, where key vertices are those that meet preset conditions and whose removal does not affect the network's critical connectivity; a subgraph building module, which constructs feasible component subgraphs from the directed graph model based on the list of key vertices and assigns adaptive weights to the edges in the feasible component subgraphs that incorporate the quality of service requirements of the healthcare business; a path finding module, which finds augmenting paths for healthcare business flows in the feasible component subgraphs based on adaptive weights, and after verification, allocates traffic along the augmenting path and updates the network state; and a reconstruction and update module, which dynamically reconstructs the network, updates the feasible component subgraphs and adaptive weights, and returns to search for augmenting paths again until the global termination condition is met when no augmenting path can be found in the feasible component subgraphs.

[0097] Working Principle: This invention abstracts the smart healthcare network into a directed graph model with service priority attributes and dynamically identifies key vertices that can maintain critical connectivity after removal. It constructs a feasible component subgraph that balances structural simplification and service assurance. Then, it finds augmenting paths based on adaptive weights that integrate healthcare service quality requirements and triggers dynamic reconstruction when a path is infeasible. On the one hand, this reduces the computational complexity of traffic allocation in large-scale healthcare networks. On the other hand, through priority awareness and dynamic reconstruction, it ensures that high-priority services such as emergency surgery control signals and emergency image transmission always obtain low-latency and highly reliable paths. This effectively solves the technical challenge of traditional maximum flow algorithms in balancing service hierarchy assurance and computational efficiency in complex healthcare scenarios.

[0098] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0099] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0100] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0101] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0102] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart healthcare network traffic allocation method based on an improved augmented path maximum flow algorithm, characterized in that, Includes the following steps: The smart healthcare network is abstracted as a directed graph model with capacity and service priority attributes; In the directed graph model, a list of key vertices is dynamically identified and managed, wherein the key vertices in the list are those that meet preset conditions and whose removal does not affect the critical connectivity of the network. Based on the list of key vertices, a feasible component subgraph is constructed from the directed graph model, and the edges in the feasible component subgraph are assigned adaptive weights that incorporate the service quality requirements of medical services. In the feasible component subgraph, augmenting paths are found for medical service flows based on the adaptive weights. After verification, traffic is allocated along the augmenting paths and the network status is updated. In response to the failure to find the augmenting path in the feasible component subgraph, the network is dynamically reconstructed, the feasible component subgraph and the adaptive weights are updated, and the search for the augmenting path is restarted until the global termination condition is met.

2. The smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm according to claim 1, characterized in that, The directed graph model includes a set of vertices, a set of directed edges, a capacity function, and a priority function; The vertex set includes a source vertex, a sink vertex, and intermediate vertices; The directed edges in the set of directed edges represent network links between vertices; The capacity function assigns a positive capacity value to each directed edge to characterize the real-time available bandwidth of the link; The priority function assigns a priority to each directed edge.

3. The smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm according to claim 1, characterized in that, The dynamic identification and management of the key vertex list includes: Traverse the set of vertices in the directed graph model and apply topology hub determination, edge protection determination, and load and service value mismatch determination to filter candidate vertices; Perform destructive connectivity verification on the candidate vertices: remove the candidate vertices and their associated edges from the memory copy, verify the connectivity from the source vertex to the critical destination by breadth-first search, and add the candidate vertex to the critical vertex list only if the connectivity is maintained; The list of critical vertices is dynamically updated based on periodic and / or event-triggered mechanisms, removing vertices that are ineffective, have low load, or have an increased proportion of high-priority traffic, and adding vertices that meet the criteria.

4. The smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm according to claim 3, characterized in that, The topology hub determination configuration is as follows: if the sum of the degrees of the vertices is greater than a preset topology hub threshold, then the vertex is marked as a candidate vertex; The edge protection determination configuration is as follows: if the in-degree or out-degree of a vertex is 0 and the service priority of its associated edge is lower than a preset protection threshold, then it is marked as the candidate vertex; The load and value mismatch determination configuration is as follows: if the load rate of a vertex is higher than the first threshold and the corresponding high-priority traffic ratio is lower than the second threshold, then it is marked as the candidate vertex.

5. The smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm according to claim 1, characterized in that, The construction of feasible component subgraphs includes: Remove vertices and their associated edges from the list of key vertices in the directed graph model to form an initial simplified subgraph; For each business flow in the set of critical business flows, verify the connectivity from the source to the destination in the initial simplified subgraph; If there are disconnected business flows, select the minimum number of vertices from the list of critical vertices and re-add them to the initial simplified subgraph to restore the connectivity of all critical business flows and obtain the feasible component subgraph.

6. The smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm according to claim 1, characterized in that, Assigning adaptive weights based on the integrated medical service quality requirements to the edges in the feasible component subgraph includes: Define a service quality requirement vector for the service flow, which includes the required thresholds for latency, bandwidth, packet loss rate, and latency jitter. For network link maintenance, a quality of service state vector is provided, which includes the estimated values ​​of corresponding latency, available bandwidth, packet loss rate, and latency jitter. The adaptive weight of the edge to the service flow is calculated by multiplying the static basic weight and the dynamic adaptive factor. The static basic weight is negatively correlated with the estimated latency of the link. The dynamic adaptive factor is obtained by multiplying the exponential adjustment terms of each service quality dimension. The exponential adjustment terms are rewarded or penalized based on the degree to which the link state value deviates from the service demand threshold, and the intensity of the reward or penalty is controlled by a temperature parameter that is dynamically adjusted according to the degree of network congestion.

7. The smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm according to claim 1, characterized in that, The process of finding augmenting paths for medical service flows based on the adaptive weights, verifying the allocation of traffic through the augmenting paths found via the trailing edges, and updating the network state includes: Based on the service quality requirement vector of the business flow to be scheduled, set corresponding adaptive weights for the edges of the feasible component subgraph; Run the minimum weighted path algorithm on the feasible component subgraph with weights set to find candidate paths from the source to the sink. Real-time probe data packets are injected into the candidate paths to measure actual latency and packet loss rate. If the business requirements are met, the path is confirmed as the selected path; otherwise, the relevant edge weights are adjusted and a new path is searched. Calculate the maximum allocatable flow for the selected path, smoothly inject the flow through the token bucket controller, and update the remaining capacity of the path edges and the vertex priority flags.

8. The smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm according to claim 1, characterized in that, The method also includes: When a newly arriving high-priority service flow has no path that meets the requirements in the feasible component subgraph and the remaining network bandwidth is below the threshold, the active low-priority service flows are traversed, and the flow that overlaps with the target path and has the lowest priority or consumes the most resources is selected. Send a pre-preemption notification to the source of the selected flow, release the bandwidth resources it occupies after receiving a response, update the feasible component subgraph, and find augmentation paths for the high-priority service flow again.

9. The smart healthcare network traffic allocation method based on the improved augmented path maximum flow algorithm according to claim 1, characterized in that, The dynamic reconfiguration includes: Scan the feasible component subgraph and remove all useless vertices and their associated edges whose remaining capacity is zero for either incoming or outgoing edges; Check the connectivity of the reconstructed feasible component subgraph. If the critical business flow is disconnected, select the vertex with the highest connectivity or the highest historical priority from the list of critical vertices and add it to the reconstructed feasible component subgraph to restore connectivity. Based on the recovered feasible component subgraph, key vertex identification, feasible component subgraph construction, and adaptive weight calculation are re-executed.

10. A smart healthcare network traffic allocation system based on an improved augmented path maximum flow algorithm, characterized in that: include: The model building module is used to abstract the smart healthcare network into a directed graph model with capacity and service priority attributes; The vertex recognition module is used to dynamically identify and manage the list of key vertices in the directed graph model, wherein the key vertices in the list of key vertices are those that meet preset conditions and whose removal does not affect the critical connectivity of the network. The subgraph construction module is used to construct feasible component subgraphs from the directed graph model based on the list of key vertices, and to assign adaptive weights to the edges in the feasible component subgraphs that incorporate the service quality requirements of medical business. The path finding module is used to find augmenting paths for medical service flows based on the adaptive weights in the feasible component subgraph, and after verification, allocate traffic along the augmenting path and update the network status. The reconstruction and update module is used to dynamically reconstruct the network in response to the failure to find the augmenting path in the feasible component subgraph, update the feasible component subgraph and the adaptive weights, and return to search for the augmenting path again until the global termination condition is met.

Images