A load balancing method for edge sensing service function chain for low-altitude IoT

Abstract

This invention discloses a load balancing method for edge-sensing service function chains in low-altitude IoT. Addressing the characteristics of high node mobility, time-varying topology, and heterogeneous resources in low-altitude IoT, this method first establishes a three-layer network model including space-based, air-based, and ground-based nodes, quantifying the constraints on node computing resources and link bandwidth resources. Then, a node segmentation technique is designed to transform node resource constraints into equivalent link capacity constraints, achieving joint optimization of computing and communication resources. Next, a service function chain scheduling algorithm based on the maximum flow minimum cut theorem is proposed, balancing computing and bandwidth resource loads through a dynamic weight adjustment mechanism. Finally, a fault recovery test program is constructed to verify the robustness of the algorithm in the event of node or link failures. Under the premise of limited resources in low-altitude IoT, this invention achieves a balanced distribution of network load by collaboratively optimizing the deployment of virtual network functions and path selection. Experimental results show that the algorithm proposed in this invention reduces peak computing resource consumption by 46.0% and peak bandwidth consumption by 19.1%. Under 40 concurrent service function chain requests, the service acceptance rate reaches 41.1%, which is 66.5% higher than the traditional algorithm. At the same time, the fault recovery rate reaches 99%, providing technical support for high-reliability services of low-altitude IoT.

Description

Technical Field

[0001] This invention belongs to the field of Internet of Things and network communication technology, specifically relating to a service function chain scheduling and load balancing method in low-altitude Internet of Things scenarios. Background Technology

[0002] With the rapid development of 5G / 6G communication technologies, drone networks, and edge computing, low-altitude IoT has become a key component of the integrated air-space-ground network. Low-altitude IoT supports a wide range of applications such as smart city management, precision agriculture, disaster response, and environmental monitoring. These applications typically rely on the orchestration and execution of virtual network functions, organized into service function chains, processing data streams in a predetermined order while meeting stringent quality of service requirements.

[0003] In the Internet of Things (IoT) environment, edge-aware intelligence enables flexible and dynamic scheduling of diverse services through service function chains. However, ensuring consistent service quality remains a significant challenge due to the heterogeneity of multi-layered resources and the inherent dynamism of networks. Low-altitude IoT architectures comprise edge nodes, air nodes, and ground nodes, which exhibit significant differences in resource availability, including the limited computing power, constrained link bandwidth, and significant cross-domain transmission latency of edge nodes. Furthermore, node mobility and frequent changes in network topology introduce additional complexity to resource allocation and service orchestration processes.

[0004] Traditional service function chain mapping and load balancing algorithms are primarily designed for static or terrestrial network infrastructure, and have significant shortcomings in highly mobile and time-varying low-altitude IoT scenarios: they often lead to uneven resource utilization, increased risk of single points of failure, and limited adaptability to rapid changes in network conditions, thereby reducing service reliability and overall resource efficiency. Specifically, existing methods have the following technical drawbacks:

[0005] (1) Separation of VNF deployment and path selection: Many traditional algorithms adopt a sequential approach, i.e. routing first and then deployment or deployment first and then routing, which leads to a mismatch between VNF deployment and path resources, resulting in local resource bottlenecks and low overall resource utilization.

[0006] (2) Single-dimensional resource optimization: Most existing load balancing algorithms focus on optimizing only one dimension of computing resources or bandwidth resources, ignoring the interdependence and constraints between multiple resources, resulting in an overall imbalance of network load.

[0007] (3) Insufficient adaptability to the heterogeneity of low-altitude IoT: Traditional models are difficult to capture the impact of differences in physical attributes, resource capacity and fault thresholds of space-based, air-based and ground-based nodes on service scheduling.

[0008] (4) Insufficient fault tolerance: Low-altitude IoT networks are susceptible to factors such as node mobility and signal interference, resulting in a high probability of node or link failures. Existing algorithms lack effective fault detection and recovery mechanisms.

[0009] Therefore, how to achieve joint optimization of computing and communication resources, balance network load, and ensure service continuity during failures in the resource-constrained, highly dynamic, and heterogeneous low-altitude IoT environment has become a pressing technical challenge. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of the prior art and provide a load balancing method for edge sensing service function chains for low-altitude Internet of Things that can achieve joint optimization of computing and communication resources, balance network load, and ensure fault recovery capability.

[0011] The solution to achieve the purpose of this invention is: a load balancing method for edge sensing service function chains for low-altitude Internet of Things, comprising the following steps:

[0012] Step 1: Establish a three-layer network model for low-altitude IoT, including space-based nodes, air-based nodes, and ground-based nodes, and define the node computing resource capacity and link bandwidth resource capacity.

[0013] Step 2: Design a node partitioning technique to transform node resource constraints into equivalent link capacity constraints;

[0014] Step 3: Propose a service function chain scheduling algorithm based on the maximum flow minimum cut theorem to achieve joint optimization of virtual network function deployment and path selection;

[0015] Step 4: Construct a dynamic weight adjustment mechanism to balance the load on computing resources and the load on bandwidth resources;

[0016] Step 5: Design a fault recovery mechanism to ensure service continuity when a node or link fails.

[0017] Furthermore, the establishment of the three-layer network model for low-altitude IoT described in step 1 is as follows:

[0018] The underlying physical network is represented as an undirected weighted graph. ,in It is a set of physical nodes, including space-based nodes, air-based nodes, and ground-based nodes; It is a set of physical links, including intra-domain links (links between nodes at the same level) and inter-domain links (links between nodes at different levels). Represents network resources, among which Calculate resource capacity for nodes. For the remaining computing resources of the node, For link bandwidth resource capacity, For the remaining bandwidth resources of the link, This refers to the link delay attribute.

[0019] To ensure the feasibility and effectiveness of service function chain mapping, the following constraints must be met:

[0020] (1) Computational resource constraints: The sum of the computational resources required by all VNFs deployed on a physical node cannot exceed the remaining computational resource capacity of that node. For any physical node ,set up For deployment in VNF sets on, For VNF Given the computing resource requirements, then:

[0021]

[0022] (2) Bandwidth resource constraint: The sum of the bandwidth requirements of all virtual links mapped to the physical link cannot exceed the remaining bandwidth capacity of that link. For any physical link ,set up For mapping to The set of virtual links on the Internet For virtual links If the bandwidth requirement is:

[0023]

[0024] (3) Sequence Constraint: The deployment order of VNFs in the service function chain must be consistent with the logical node order of the service chain. For any two consecutive VNFs in an SFC... and ,deploy The physical nodes must be able to access the deployment via physical links. The physical nodes are reachable.

[0025] (4) Transmission delay constraint: The total transmission delay of the SFC path, including the processing delay of the VNF and the transmission delay of the link, must not exceed the user's maximum delay requirement. Let... For VNF Processing latency, For physical links Transmission delay, If the user's maximum latency requirement is met, then:

[0026]

[0027] in This refers to the set of physical links in the SFC path.

[0028] Furthermore, the node segmentation technique described in step 2 is as follows:

[0029] To transform node resource constraints into equivalent link capacity constraints, each physical node is partitioned. For each physical node... Divide it into entry nodes and export node ,exist and A directed edge is introduced between them, with its capacity set to the computing resource capacity of that node, i.e. Remove all original links from the node. To the node The edge is updated from arrive The connection maintains the original link capacity. constant.

[0030] Through the above partitioning, node resource constraints are transformed into equivalent link capacity constraints, enabling the maximum flow problem to be solved in the modified graph, laying the foundation for subsequent joint optimization.

[0031] Furthermore, the service function chain scheduling algorithm based on the maximum flow minimum cut theorem described in step 3 specifically includes the following steps:

[0032] First, all service function chain requests are sorted in descending order of computing and bandwidth resource requirements, with priority given to requests with high resource requirements, in order to improve overall resource utilization efficiency.

[0033] For each request, the following process is used for scheduling:

[0034] (1) Maximum Flow Calculation: For each arriving SFC request, construct a network containing partitioned nodes, and use the maximum flow algorithm to calculate the maximum flow from the source node to the target node. Iteratively search for the maximum flow from the source node using breadth-first search. to the destination node The system first identifies augmenting paths, then uses a depth-first search to traverse these paths to increase the traffic flowing out of the source node. If an identified path satisfies capacity constraints, the traffic is increased accordingly, and the remaining capacity is updated during the backtracking process. The entire process ensures that the traffic on each path does not exceed the resource limits of any node or the bandwidth constraints of the link.

[0035] (2) Minimum Cut Calculation: For a given maximum flow, the capacity of the minimum cut is calculated using the minimum cut theorem. This capacity represents the bottleneck of the maximum flow. According to the Ford-Fulkerson theorem, the value of the maximum flow is equal to the capacity of the minimum cut. By identifying nodes or edges that act as traffic bottlenecks, targeted traffic adjustments or network configuration optimizations can be implemented. By adjusting the deployment of SFCs and optimizing path selection, resource utilization can be improved while reducing the risk of node and link overload.

[0036] Furthermore, the dynamic weight adjustment mechanism described in step 4 is as follows:

[0037] To achieve joint optimization of computing and bandwidth resources, two key metrics are defined: maximum normalized computing resource consumption. and maximum normalized bandwidth resource consumption .

[0038] The maximum normalized computational resource consumption is defined as:

[0039]

[0040] in For the remaining computing resources of the node, Calculate the resource capacity for the node.

[0041] The maximum normalized bandwidth resource consumption is defined as:

[0042]

[0043] in This represents the remaining bandwidth of the link. This refers to the link bandwidth capacity.

[0044] To consider both factors simultaneously, a weighted average and a specific technique are used, and the optimization objective function is set as follows:

[0045]

[0046] in These are weighting coefficients used to balance the importance of computational resource load and bandwidth resource load. The objective function considers not only the load on nodes and links, but also the request acceptance rate, which is crucial for load balancing.

[0047] To comprehensively evaluate algorithm performance, a weighted score based on comprehensive resource consumption is introduced:

[0048]

[0049] Based on the experimental results, the dynamic weight adjustment mechanism uses computational resource weights. Bandwidth resource weight This improves the overall load score by 32% compared to the static strategy.

[0050] Furthermore, the fault recovery mechanism described in step 5 is as follows:

[0051] To verify the robustness of the LB-SFC algorithm, a fault recovery test procedure was designed. The experiment employed a three-layer IoT topology with 100 nodes, including 10 satellite nodes, 25 airborne nodes, and 65 ground nodes. Transient fault injection was implemented by randomly selecting 5% of the physical nodes (5 nodes) and 5% of the physical links (26 links) to simulate network component failure scenarios.

[0052] The following key metrics will be continuously monitored during the testing process:

[0053] (1) Rerouting time: The average time from the occurrence of the failure to the completion of path reconstruction for all affected SFC requests;

[0054] (2) Request recovery rate: The percentage of SFC requests that were successfully recovered out of the total number of affected requests;

[0055] (3) Resource overhead: The proportion of additional computing and bandwidth resources consumed during the recovery process;

[0056] (4) Quality of Service Assurance: Success rate of recovery for high-priority requests (accounting for 25% of total requests).

[0057] The LB-SFC algorithm quickly identifies and avoids faulty regions by calculating the maximum flow and minimum cut of the network in real time. It utilizes node segmentation technology to transform computational resource constraints into link capacity constraints, achieving cross-layer resource collaborative optimization. An incremental update algorithm is employed to locally adjust affected network regions, avoiding the performance bottleneck of traditional algorithms' full-scale computation model. Simultaneously, a priority-aware recovery strategy is designed to ensure that critical services are migrated first, maintaining a recovery rate of nearly 99% even in complex fault scenarios. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the three-layer network collaborative control architecture for low-altitude Internet of Things in this invention.

[0059] Figure 2 This is a diagram illustrating the load balancing selection process in this invention.

[0060] Figure 3 This is a flowchart of the LB-SFC service function chain scheduling algorithm in this invention. Detailed Implementation

[0061] The present invention will now be described in detail with reference to the accompanying drawings.

[0062] This invention discloses an edge sensing service function chain load balancing method (LB-SFC) for low-altitude Internet of Things, comprising the following steps:

[0063] Step 1: Establish a three-layer network model for low-altitude Internet of Things (IoT).

[0064] Combination Figure 1 The low-altitude IoT network architecture is divided into three layers: space-based, air-based, and ground-based. Space-based nodes include low-Earth orbit satellites, medium-Earth orbit satellites, and high-Earth orbit satellites, providing wide-area coverage; air-based nodes include drones and high-altitude airships, serving as relay hubs between space-based and ground-based nodes; and ground-based nodes include 5G base stations, core network equipment, and end-user equipment, providing stable service access and data processing capabilities.

[0065] The underlying physical network is represented as an undirected weighted graph. Resource attributes include node computing resource capacity, remaining computing resources, link bandwidth capacity, remaining bandwidth resources, and link latency. Set the resource parameter ranges for each layer of nodes and links according to Table 1, and set the parameter ranges for service function chain requests according to Table 2. These must meet computing resource constraints, bandwidth resource constraints, and transmission latency constraints.

[0066] Step 2: Design node segmentation technology.

[0067] To transform node resource constraints into equivalent link capacity constraints, each physical node is partitioned. For each physical node... Divide it into entry nodes and export node ,exist and A directed edge is introduced between them, with its capacity set to the computing resource capacity of that node, i.e. Remove all original links from the node. To the node The edge is updated from arrive The connections are maintained while preserving the original link capacity. By partitioning nodes, the maximum flow problem can be solved in the modified graph.

[0068] Step 3: Service function chain scheduling algorithm based on the maximum flow minimum cut theorem.

[0069] First, all service function chain requests are sorted in descending order of computational and bandwidth resource requirements. For each request, a set of feasible paths from the source node to the target node is obtained through maximum flow computation, ensuring that each path meets node resource constraints and link bandwidth constraints. Nodes or edges acting as traffic bottlenecks are identified through minimum cut computation, and targeted traffic adjustments are implemented. Algorithms 1 and 2 provide the complete scheduling process.

[0070] Step 4: Dynamic weight adjustment mechanism.

[0071] Define the maximum normalized computational resource consumption (Equation 4) and the maximum normalized bandwidth resource consumption (Equation 5), construct the optimization objective function (Equation 6), and balance the computational resource load and bandwidth resource load through dynamic weight adjustment. Computational resource weights are used. Bandwidth resource weight This improves the overall load score by 32% compared to the static strategy.

[0072] Step 5: Fault recovery mechanism.

[0073] Transient fault injection is implemented by randomly selecting 5% of physical nodes and 5% of physical links to simulate network component failure scenarios. Four core metrics are monitored: rerouting time, request recovery rate, resource overhead, and quality of service assurance. The LB-SFC algorithm quickly identifies and avoids faulty areas by calculating the maximum flow and minimum cut in real time, and utilizes node segmentation technology to achieve cross-layer resource collaborative optimization, achieving a fault recovery rate of 99%.

[0074] This invention patent proposes a load balancing method for edge sensing service function chains in low-altitude IoT. The algorithm pseudocode can be defined as follows:

[0075]

Claims

1. A load balancing method for edge sensing service function chains in low-altitude Internet of Things (IoT), characterized in that, Includes the following steps: Step 1: Establish a three-layer network model for low-altitude IoT, including space-based nodes, air-based nodes, and ground-based nodes, and define the node computing resource capacity and link bandwidth resource capacity. Step 2: Design a node partitioning technique to transform node resource constraints into equivalent link capacity constraints; Step 3: Propose a service function chain scheduling algorithm based on the maximum flow minimum cut theorem to achieve joint optimization of virtual network function deployment and path selection; Step 4: Construct a dynamic weight adjustment mechanism to balance the load on computing resources and the load on bandwidth resources; Step 5: Design a fault recovery mechanism to ensure service continuity when a node or link fails.

2. The edge sensing service function chain load balancing method for low-altitude Internet of Things according to claim 1, characterized in that, Step 1 involves establishing a three-layer network model for low-altitude IoT, as detailed below: The underlying physical network is represented as an undirected weighted graph. ,in It is a set of physical nodes, including space-based nodes, air-based nodes, and ground-based nodes; It is a set of physical links, including intra-domain links and inter-domain links; Represents network resources, among which Calculate resource capacity for nodes. For the remaining computing resources of the node, For link bandwidth resource capacity, For the remaining bandwidth resources of the link, This refers to the link delay attribute; The computing resource capacity of the space-based nodes is set to follow a uniform distribution. Empty base nodes obey Foundation nodes obey Space-based link bandwidth follows Airborne links obey Ground-based links obey Inter-domain links obey Space-based link delay follows Milliseconds, air-based link obeys Milliseconds, ground-based link obeys Milliseconds, inter-domain links obey millisecond.

3. The edge sensing service function chain load balancing method for low-altitude Internet of Things according to claim 1, characterized in that, The node segmentation technique described in step 2 is as follows: For each physical node Divide it into entry nodes and export node ; exist and A directed edge is introduced between them, with its capacity set to the computing resource capacity of that node, i.e. ; All raw links from node To the node The edge is updated from arrive The connection maintains the original link capacity. constant; By performing the above segmentation, the node resource constraint is transformed into an equivalent link capacity constraint, enabling the maximum flow problem to be solved in the modified graph.

4. The edge sensing service function chain load balancing method for low-altitude Internet of Things according to claim 1, characterized in that, Step 3, the service function chain scheduling algorithm based on the maximum flow minimum cut theorem, specifically includes the following steps: First, sort all service function chain requests in descending order of computing resource and bandwidth resource requirements; For each request, a breadth-first search is used to construct a hierarchical graph to find an augmenting path from the source node to the destination node; A depth-first search is used to traverse augmenting paths, increasing the traffic from the source node to the destination node, and ensuring that the traffic on each path does not exceed the node resource limit and link bandwidth constraint. According to the Ford-Fulkerson theorem, the maximum flow is equal to the capacity of the minimum cut. By identifying nodes or edges that act as flow bottlenecks, targeted flow adjustments can be implemented.

5. The edge sensing service function chain load balancing method for low-altitude Internet of Things according to claim 1, characterized in that, The dynamic weight adjustment mechanism described in step 4 is as follows: Define the maximum normalized computational resource consumption. and maximum normalized bandwidth resource consumption : ； in For the remaining computing resources of the node, Calculate resource capacity for nodes. This represents the remaining bandwidth of the link. This refers to the link bandwidth capacity. The objective function is set as follows: ； in These are weighting coefficients used to balance the importance of computational resource load and bandwidth resource load. Request acceptance rate; The dynamic adjustment mechanism updates the weight coefficients in real time based on the network status and calculates the resource weights. Bandwidth resource weight This improves the overall load score by 32% compared to the static strategy.

6. The edge sensing service function chain load balancing method for low-altitude Internet of Things according to claim 1, characterized in that, The fault recovery mechanism described in step 5 is as follows: Transient fault injection was performed by randomly selecting 5% of physical nodes and 5% of physical links to simulate network component failure scenarios. Monitor the following key metrics: (1) Rerouting time: The average time from the occurrence of a failure to the completion of path reconstruction for all affected service function chain requests; (2) Request recovery rate: The percentage of service function chain requests that were successfully recovered out of the total number of affected requests; (3) Resource overhead: The proportion of additional computing and bandwidth resources consumed during the recovery process; (4) Service Quality Assurance: Success rate of recovery for high-priority requests (accounting for 25% of total requests); The algorithm quickly identifies and avoids faulty regions by calculating the maximum flow and minimum cut of the network in real time. It uses node segmentation technology to transform computational resource constraints into link capacity constraints, achieving cross-layer resource collaborative optimization and a fault recovery rate of 99%. This invention patent proposes a load balancing method for edge sensing service function chains in low-altitude IoT. The algorithm pseudocode can be defined as follows: 。

Images