A dynamic migration method for edge computing tasks supporting service continuity

Abstract

The present invention discloses a method for dynamic migration of edge computing tasks that supports service continuity, comprising anchoring the original edge node on which the computing task runs, and marking M first candidate nodes to which the computing task is to be migrated; obtaining a static attribute vector corresponding to a node index number; screening Q second candidate nodes from the M first candidate nodes; predicting several dynamic operating parameters of the computing task at the Q second candidate nodes; calculating the dynamic operating quality and Euclidean distance of the Q second candidate nodes; generating a distance node sequence and a quality node sequence based on the Euclidean distance and the dynamic operating quality; aligning the distance node sequence and the quality node sequence based on descending numbering to generate a node pair sequence, and determining the optimal migration node in the node pair sequence; migrating the computing task to the optimal migration node; the present invention realizes the dual condition fusion of migration node determination, thereby ensuring the service continuity of the computing task during the dynamic migration process.

Description

Technical Field

[0001] The present invention relates to the dynamic migration of edge tasks, and specifically to a method for dynamic migration of edge computing tasks that supports service continuity. Background Art

[0002] With the widespread deployment of edge computing technology, low-latency, high-bandwidth collaboration between terminal devices and edge nodes has become a key support for ensuring service quality. In actual operation, edge nodes may experience performance degradation due to factors such as resource load fluctuations, network degradation, or hardware degradation, affecting the stable execution of computing tasks and causing service interruptions or response delays. Therefore, how to achieve dynamic migration of computing tasks while ensuring service continuity is a core technical challenge in edge computing systems.

[0003] Existing edge computing task migration schemes primarily use a single metric to determine migration nodes. One approach ranks candidate nodes based on their operational quality and selects nodes with higher scores for task migration. While this approach can improve computing performance to a certain extent, it lacks the ability to assess the relative relationship between a node's operational status and the original degraded node. This can result in tasks being migrated to nodes that appear to have higher quality but are still close to degraded regions, impacting the effectiveness of migration. Another approach uses the difference in status between a node and a degraded node. For example, distance metrics are constructed using communication latency, network packet loss rate, and operational anomaly indicators, selecting the node with the greatest state distance as the migration target. This approach can avoid degraded regions to a certain extent, but without a comprehensive assessment of node computing power and service quality, it can result in tasks being migrated to resource-constrained nodes, thereby reducing overall service performance.

[0004] In summary, existing technologies still have the following problems in edge computing task migration: there is a lack of unified consideration of "node operation quality" and "node status difference", making it difficult to balance the dual goals of service performance and degradation avoidance; especially when the sorting results are inconsistent, there is a lack of effective decision-making basis to determine the optimal migration node. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a dynamic migration method for edge computing tasks that supports service continuity. By determining the optimal migration node through dual-sequence fusion alignment, the optimal migration node can be screened under dual constraints. At the same time, by introducing a shared node index number and a third candidate node fallback, the technical problems raised in the background technology are solved.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] A method for dynamically migrating edge computing tasks to support service continuity includes:

[0008] S1. Anchor the original edge node where the computing task is running, and mark M first candidate nodes to which the computing task is to be migrated;

[0009] S2. Assign unique node index numbers to the M first candidate nodes and the original edge node, and obtain the static attribute vectors corresponding to the node index numbers;

[0010] S3. Screen Q second candidate nodes from the M first candidate nodes according to the static attribute vector corresponding to the node index number;

[0011] S4. Concatenate the static attribute vector of the second candidate node and the standardized features of the computing task as joint inputs of a parameter regression model to predict several dynamic operating parameters of the computing task on the Q second candidate nodes;

[0012] S5. Calculating the dynamic operation qualities of the Q second candidate nodes based on the predicted dynamic operation parameters, and calculating the Euclidean distances between the second candidate nodes and the original edge nodes;

[0013] S6. Arrange the Q second candidate nodes in descending order according to the Euclidean distance and the dynamic operation quality to generate a distance node sequence and a quality node sequence;

[0014] The distance node sequence and the quality node sequence are both arranged in descending order, and each second candidate node is assigned a descending number from the front to the back;

[0015] S7, aligning the distance node sequence and the quality node sequence based on the descending numbering, generating a node pair sequence, and determining the optimal migration node in the node pair sequence;

[0016] The node pair sequence includes Q binary node pairs arranged in descending order, and each binary node pair is paired with two second candidate nodes from the distance node sequence and the quality node sequence;

[0017] S8. Migrate the computing task to the optimal migration node.

[0018] In some specific embodiments, marking the M first candidate nodes to which the computing task is to be migrated includes:

[0019] S1-1. Obtaining the dynamic running quality of the computing task on the original edge node;

[0020] S1-2. Determine whether the original edge node is a degenerate node based on the dynamic operation quality;

[0021] S1-3. If it is a degenerate node, mark the M first candidate nodes to which the computing task is to be migrated; otherwise, continue to run the computing task on the original edge node.

[0022] In some specific embodiments, obtaining the dynamic running quality of the computing task on the original edge node includes:

[0023] S1-1-1. Slidingly collect several dynamic operating parameters of the computing task at the original edge node within a continuous time window;

[0024] S1-1-2. Calculate the dynamic operation quality of the computing task within the time window based on several dynamic operation parameters of the original edge node.

[0025] In some specific embodiments, assigning unique node index numbers to the M first candidate nodes and the original edge node includes:

[0026] S2-1, mark several edge nodes that have network connections with the original edge node;

[0027] S2-2. Calculate the spatial distances between several edge nodes and the original edge node;

[0028] S2-3, marking the edge node whose spatial distance is lower than the threshold as the first candidate node;

[0029] S2-4. Assign unique node index numbers to the first candidate node and the original edge node according to the spatial distance between the first candidate node and the original edge node.

[0030] In some specific embodiments, screening Q second candidate nodes from M first candidate nodes according to the static attribute vector corresponding to the node index number includes:

[0031] S3-1. Perform K-nearest-neighbor clustering on the static attribute vectors of the M first candidate nodes to obtain K static clusters;

[0032] S3-2, average pooling is performed on several static attribute vectors within the K static clusters to generate K cluster center vectors corresponding to the static clusters;

[0033] S3-3, obtaining a static attribute vector of the original edge node, and calculating a first similarity between the static attribute vector and the K cluster center vectors;

[0034] S3-4. Sort the K static clusters corresponding to the cluster center vectors according to the K first similarities to generate a cluster sequence;

[0035] S3-5, starting from the static cluster corresponding to the maximum first similarity, intercepting the first Q similar clusters in the cluster sequence;

[0036] S3-6, extract the cluster center vectors of the first Q similar clusters from the K cluster center vectors, and anchor the static attribute vector closest to the cluster center vector in each similar cluster;

[0037] S3-7. Obtain the nearest static attribute vector and its node index number, and select Q second candidate nodes based on the node index number;

[0038] In some specific embodiments, calculating the Euclidean distance between the second candidate node and the original edge node includes:

[0039] S5-1. Mark the timestamp of when the original edge node is determined to be a degenerate node on the time axis;

[0040] S5-2. Mark the dynamic operating parameters of the original edge node at the timestamp as abnormal operating parameters;

[0041] S5-3. Normalize the features of several abnormal operation parameters of the original edge node and combine them into an abnormal operation vector of the original edge node;

[0042] S5-4. Normalize the characteristics of several dynamic operation parameters of the Q second candidate nodes and concatenate them into dynamic operation vectors of the Q second candidate nodes;

[0043] S5-5. Calculate Q Euclidean distances between the dynamic operation vector and the abnormal operation vector.

[0044] In some specific embodiments, determining the optimal migration node in a sequence of node pairs includes:

[0045] S7-1. Determine whether there is a binary node pair sharing a node index number in the node pair sequence;

[0046] S7-2. If there is a binary node pair that shares a node index number, mark the second candidate node of the binary node pair as the optimal migration node and terminate the search; otherwise, traverse the node pair sequence backward along the descending order of the binary node pairs;

[0047] S7-3. If there is no binary node pair that shares a node index number, define the two second candidate nodes in the binary node pair corresponding to the largest descending number as the third candidate nodes;

[0048] S7-4. Determine the optimal migration node based on the third candidate node.

[0049] In some specific embodiments, determining whether there is a binary node pair sharing a node index number in the node pair sequence includes:

[0050] S7-1-1. In the node pair sequence, associate the node index numbers of the two second candidate nodes of each binary node pair;

[0051] S7-1-2, search in descending order along the binary node pairs, and compare the node index numbers of the two second candidate nodes;

[0052] S7-1-3. If the node index numbers of the two second candidate nodes are the same, the corresponding binary node pair is determined to be a binary node pair that shares the node index number.

[0053] In some specific embodiments, searching for an optimal migration node based on the third candidate node includes:

[0054] S7-4-1. Obtain the node index number of the third candidate node, and extract the static attribute vectors of the two third candidate nodes based on the node index number;

[0055] S7-4-2. Calculate the second similarity between the static attribute vectors of the two third candidate nodes and the original edge node vector;

[0056] S7-4-3. Determine the second candidate node with the greater second similarity as the optimal migration node.

[0057] The present invention provides a method for dynamic migration of edge computing tasks that supports service continuity, which has the following beneficial effects:

[0058] In summary, the present invention constructs a quality node sequence and a distance node sequence by numbering the dynamic operation quality of the second candidate node and the Euclidean distance between the second candidate node and the abnormal operation state of the original edge node in descending order, and generates a node pair sequence accordingly. On this basis, if there is a second candidate node with a shared node index number in the node pair, it can be directly determined as the optimal migration node; if it is not hit, the static attribute similarity of the two second candidate nodes corresponding to the largest descending number in the node pair sequence is supplemented to determine the path, so that when the sorting results are inconsistent, the migration target with the static attribute closest to the original node can still be screened out. The present invention realizes the dual condition fusion of migration node determination, ensuring the service continuity of computing tasks during dynamic migration. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A flowchart of a method for dynamic migration of edge computing tasks supporting service continuity according to the present invention;

[0060] Figure 2 Schematic diagram of the screening process of the second candidate node of the present invention;

[0061] Figure 3 Schematic diagram of the process of determining the optimal migration node according to the present invention;

[0062] Figure 4 This is a schematic diagram of the definition process of the third candidate node according to the present invention. DETAILED DESCRIPTION

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0064] See also Figures 1 to 4 The present invention provides a method for dynamic migration of edge computing tasks supporting service continuity, comprising the following steps:

[0065] S1. Anchor the original edge node where the computing task is running, and mark M first candidate nodes to which the computing task is to be migrated.

[0066] Exemplarily, step S1 specifically includes:

[0067] S1-1. Obtaining the dynamic running quality of the computing task on the original edge node;

[0068] S1-2. Determine whether the original edge node is a degenerate node based on the dynamic operation quality;

[0069] S1-3. If it is a degenerate node, mark the M first candidate nodes to which the computing task is to be migrated; otherwise, continue to run the computing task on the original edge node.

[0070] Specifically, the dynamic operation quality may be compared with a dynamic operation quality threshold; if the dynamic operation quality is lower than the dynamic operation quality threshold, the original edge node is determined to be a degenerate node, and the computing task is marked as a task to be migrated.

[0071] Furthermore, step S1-1 specifically includes:

[0072] S1-1-1. Slidingly collect several dynamic operating parameters of the computing task at the original edge node within a continuous time window;

[0073] Among them, dynamic operation parameters include: average response time, average network delay, network packet loss rate, CPU usage and memory usage;

[0074] For example, edge nodes are computing resource units deployed close to data sources or end users, equipped with basic computing, storage, and communication capabilities, and capable of independently handling some tasks. Examples include edge servers, smart gateways, and industrial control terminals.

[0075] Furthermore, the dynamic operation parameters can reflect the dynamic operation and network status of the edge computing device in terms of task execution efficiency, network communication quality, and resource occupancy level.

[0076] Specifically, the time window slides using a unit time step to ensure that each slide triggers a real-time update of the dynamic parameters, that is, each time the time window slides a unit step, parameters such as average response time and average network delay need to be re-collected and calculated.

[0077] S1-1-2. Calculate the dynamic operation quality of the computing task within the time window based on several dynamic operation parameters at the original edge node;

[0078] The calculation formula of the dynamic running quality is:

[0079]

[0080] in, is the dynamic operation quality of the original edge node, and its value range is ,characterizes the dynamic operation of the task; For the network sub-item, For resource sub-items; is the average response time, which is used to indicate the time taken to process the task; is the average network delay, which is used to represent the communication time between original edge nodes; is the packet loss rate, which is used to reflect the instability of the network; is a very small positive number (such as ), used to prevent the denominator from being 0; is the CPU usage, which indicates the CPU resource usage of the original edge node; is the memory usage, which indicates the memory usage of the original edge node; Adjust parameters for network sub-items to control the sensitivity of network sub-items. Adjust parameters for resource sub-items to control the sensitivity of resource sub-items;

[0081] Specifically, Used to comprehensively measure task delay and network quality; where the product Indicates that the original edge node has high response time and high network delay during task processing, and the penalty should be increased; further, the denominator Indicates the packet loss rate When it increases, the denominator decreases and the network sub-item increases, thus simulating the phenomenon of performance degradation caused by high packet loss rate. It is used to reflect the resource bottleneck synergy of the original edge node; where the product This means that only when both resources are simultaneously highly utilized will the resource sub-item be significantly amplified, resulting in a significant quality penalty. For example, if both are 0.8, the product is 0.64; however, if one is 0.8 and the other is only 0.1, the product is only 0.08. The former is eight times the latter, reflecting the system's sensitivity to collaborative bottlenecks. The product is multiplied by 1 to ensure that the resource sub-item is ≥ 1 and to avoid negative exponents.

[0082] This formula uses an exponential decay structure to comprehensively evaluate the impact of average response time, average network latency, network packet loss rate, CPU utilization, and memory utilization on the dynamic operation of the origin edge node. The formula's functional structure exhibits nonlinear amplification properties, significantly converting increases in these parameters into decreases in the dynamic operation quality (Q).

[0083] The method further comprises:

[0084] S2. Assign unique node index numbers to the M first candidate nodes and the original edge node, and obtain static attribute vectors corresponding to the node index numbers.

[0085] Specifically, the static attribute vector is composed of several static attribute features of the first candidate node and the original edge node, including: geographic location information, computing power, network access bandwidth, etc., and remains relatively unchanged during task execution.

[0086] Exemplarily, step S2 includes:

[0087] S2-1, mark several edge nodes that have network connections with the original edge node;

[0088] S2-2. Calculate the spatial distances between several edge nodes and the original edge node;

[0089] S2-3, marking the edge node whose spatial distance is lower than the threshold as the first candidate node;

[0090] S2-4. Assign unique node index numbers to the first candidate node and the original edge node according to the spatial distance between the first candidate node and the original edge node.

[0091] The node index number is encoded according to the spatial distance. For example, the closer the distance, the higher the index value (or the smaller it is) to facilitate sorting, mapping, and alignment in subsequent operations.

[0092] It should be noted that the spatial distance between the original edge node and itself is defined as a null value; when assigning node index numbers to all participating nodes, ascending spatial distance encoding can be used to ensure that the node index can simultaneously reflect the physical topology structure.

[0093] The method further comprises:

[0094] S3. Screen Q second candidate nodes from the M first candidate nodes according to the static attribute vector corresponding to the node index number.

[0095] Exemplarily, step S3 specifically includes:

[0096] S3-1. Perform K-nearest-neighbor clustering on the static attribute vectors of the M first candidate nodes to obtain K static clusters;

[0097] It should be noted that each static cluster in this embodiment includes multiple static attribute vectors, which are multidimensional vectors composed of static features generated by multiple static attribute parameters of corresponding edge nodes after outlier processing and normalization.

[0098] Therefore, the “static features” described in this embodiment are all standard vectors with unified structure and consistent dimensions, which can accurately reflect the attribute characteristics of each node in static dimensions such as computing power, network access, and geographical location.

[0099] S3-2, average pooling is performed on several static attribute vectors within the K static clusters to generate K cluster center vectors corresponding to the static clusters;

[0100] It should be noted that the static attribute vectors within each static cluster have N-dimensional features. Since static attribute vectors in K-nearest neighbor clustering typically exhibit high consistency across all dimensions, they can be considered close to each other in the multidimensional feature space. However, to generate a representative vector for each static cluster, a unified representation of the static attribute vectors within the cluster is still required. Because the vectors within a cluster are relatively close in each dimension, an average pooling operation can be performed on them to generate a representative cluster centroid vector, which is used to summarize the overall static attributes of the cluster.

[0101] S3-3, obtaining a static attribute vector of the original edge node, and calculating a first similarity between the static attribute vector and the K cluster center vectors;

[0102] Specifically, the first similarity can be selected from a variety of vector similarity metrics, such as cosine similarity, the negative value of Euclidean distance, the inverse of Manhattan distance, etc.

[0103] S3-4. Sort the K static clusters corresponding to the cluster center vectors according to the K first similarities to generate a cluster sequence;

[0104] S3-5, starting from the static cluster corresponding to the maximum first similarity, intercepting the first Q similar clusters in the cluster sequence;

[0105] Specifically, in this embodiment, the value range of Q is negatively correlated with the number of clusters K. Since a larger number of clusters K indicates a more discrete distribution of node static features, and a higher probability of clusters that closely match the static features of the original edge nodes, the value range of Q can be appropriately narrowed.

[0106] For example, Q can be set by the following empirical rule:

[0107] If K≤10, then Q=0.4K;

[0108] If 10<K<30, then Q=0.2K;

[0109] If K>30, then Q=0.1K;

[0110] S3-6, extract the cluster center vectors of the first Q similar clusters from the K cluster center vectors, and anchor the static attribute vector closest to the cluster center vector in each similar cluster;

[0111] Specifically, the nearest static vector refers to a static attribute vector with the smallest Euclidean distance to the cluster center vector in similar clusters.

[0112] It should be noted that the cluster center vector is the average pooling result of the cluster center, so it may not strictly correspond to a real edge node, so the nearest static attribute vector is needed to represent the real edge node.

[0113] S3-7. Obtain the nearest static attribute vector and its node index number, and select Q second candidate nodes based on the node index number;

[0114] Specifically, each nearest static attribute vector is uniquely associated with a first candidate node through its one-to-one node index. Thus, the first candidate node corresponding to the nearest static attribute vector can be clearly selected as the second candidate node, forming a mapping path of "cluster center → nearest static attribute vector → node index → ​​second candidate node."

[0115] The method further comprises:

[0116] S4. Concatenate the static attribute vector of the second candidate node and the standardized features of the computing task as the joint input of the parameter regression model to predict several dynamic operating parameters of the computing task on the Q second candidate nodes.

[0117] It should be noted that the standardized features of the computing task include: task type, input data scale, number of threads, response time constraint, interaction frequency, etc., which have a unified dimension and scale after normalization or standardization. The standardized features of the computing task are then spliced ​​with the static attribute vector of each second candidate node as the joint input of the parameter model.

[0118] Furthermore, the parameter regression model may be a common regression prediction model such as support vector regression (SVR) or multi-layer perceptron (MLP). The regression prediction model is trained offline using historical training samples.

[0119] For example, historical training samples are constructed as follows:

[0120] First, the historical computing tasks are standardized according to the preset semantics or target features;

[0121] Then, the standardized features of each historical task are concatenated with the static attribute vector corresponding to the edge node (i.e., the second candidate node) on which it actually ran to generate a task-node joint feature representation, which is used to characterize the matching relationship between the computing task and the edge node in terms of static attributes.

[0122] Next, using this joint feature representation as input, several dynamic operating parameters (such as response time, network latency, packet loss rate, CPU and memory usage) collected during the actual operation of the historical computing task on the corresponding node are used as multi-target labels to form supervised learning samples for the regression model.

[0123] Then, the model performance is optimized through loss functions such as mean square error (MSE), and strategies such as cross-validation are used to control training error and generalization ability. Finally, based on several historical training samples, the mapping relationship between computing tasks and node attributes to dynamic operation parameters is iteratively fitted.

[0124] The method further comprises:

[0125] S5. Calculate the dynamic operation qualities of the Q second candidate nodes according to the predicted dynamic operation parameters, and calculate the Euclidean distances between the second candidate nodes and the original edge nodes.

[0126] Exemplarily, step S5 specifically includes:

[0127] S5-1. Mark the timestamp of when the original edge node is determined to be a degenerate node on the time axis;

[0128] S5-2. Mark the dynamic operating parameters of the original edge node at the timestamp as abnormal operating parameters;

[0129] It should be noted that "abnormal operating parameters" refers to the set of dynamic operating parameters collected by edge nodes at the moment a computing task is identified as a candidate for migration. This set has a clear timestamp (such as the central timestamp of the time window or the front-end and back-end timestamps). Because the computing task corresponding to this timestamp has been identified as a candidate for migration, all dynamic operating parameters at this timestamp are uniformly marked as abnormal. The specific cause of the abnormality may be a single parameter abnormality or the combined effect of multiple parameters.

[0130] S5-3. Normalize the features of several abnormal operation parameters of the original edge node and combine them into an abnormal operation vector of the original edge node;

[0131] S5-4. Normalize the characteristics of several dynamic operation parameters of the Q second candidate nodes and concatenate them into dynamic operation vectors of the Q second candidate nodes;

[0132] S5-5. Calculate Q Euclidean distances between the dynamic operation vector and the abnormal operation vector.

[0133] This embodiment achieves quantitative representation of the difference in operating states by uniformly normalizing the abnormal operating parameters and the operating parameters of the second candidate node and calculating their Euclidean distance.

[0134] The method further comprises:

[0135] S6. Arrange the Q second candidate nodes in descending order according to the Euclidean distance and the dynamic operation quality to generate a distance node sequence and a quality node sequence.

[0136] The distance node sequence and the quality node sequence are both arranged in descending order, and a descending number is assigned to each second candidate node from the front to the back.

[0137] In the distance node sequence, after sorting the Euclidean distance values ​​from large to small, each second candidate node is assigned a descending number in sequence, from Q to 1. In the quality node sequence, after sorting the dynamic operation quality from large to small, the nodes are also assigned a descending number in sequence, from Q to 1.

[0138] The method further comprises:

[0139] S7. Align the distance node sequence and the quality node sequence based on the descending numbering to generate a node pair sequence, and determine the optimal migration node in the node pair sequence.

[0140] The node pair sequence includes Q binary node pairs arranged in descending order, and each binary node pair is paired with two second candidate nodes from the distance node sequence and the quality node sequence.

[0141] Exemplarily, step S7 specifically includes:

[0142] S7-1. Determine whether there is a binary node pair sharing a node index number in the node pair sequence;

[0143] S7-3. If there is a binary node pair that shares a node index number, mark the second candidate node of the binary node pair as the optimal migration node and terminate the search; otherwise, traverse the node pair sequence backward along the descending order of the binary node pairs;

[0144] S7-4. If there is no binary node pair that shares a node index number, define the two second candidate nodes in the binary node pair corresponding to the largest descending number as the third candidate nodes;

[0145] Specifically, the maximum descending number refers to the node pair with the highest ranking in the sequence. This refers to the node pair whose two nodes rank first (descending number Q) in both the distance node sequence and the quality node sequence. This node pair performs optimally in both evaluation dimensions. Although they do not hit the same node, they can be considered a representative node with high operational quality and excellent operating status.

[0146] S7-4. Determine the optimal migration node based on the third candidate node.

[0147] This embodiment implements dual-path determination of the optimal migration node by setting a matching path that shares a node index number and a fallback comparison path based on the third candidate node.

[0148] Furthermore, step S7-2 specifically includes:

[0149] S7-2-1. In the node pair sequence, associate the node index numbers of the two second candidate nodes of each binary node pair;

[0150] S7-2-2, search in descending order along the binary node pairs, and compare the node index numbers of the two second candidate nodes;

[0151] S7-2-3. If the node index numbers of the two second candidate nodes are the same, the corresponding binary node pair is determined to be a binary node pair that shares the node index number.

[0152] Specifically, if the node index numbers are the same, that is, the corresponding second candidate nodes in the distance node sequence and the quality node sequence point to the same edge node, it indicates that the second candidate node ranks best in both operating status (long Euclidean distance) and operating quality (good performance), and can be regarded as the optimal migration node with double hits.

[0153] Exemplarily, step S7-4 specifically includes:

[0154] S7-4-1. Obtain the node index number of the third candidate node, and extract the static attribute vectors of the two third candidate nodes based on the node index number;

[0155] S7-4-2. Calculate the second similarity between the static attribute vectors of the two third candidate nodes and the original edge node vector;

[0156] Exemplarily, the second similarity preferably uses the same similarity calculation strategy as the first similarity.

[0157] S7-4-3. Determine the second candidate node with the greater second similarity as the optimal migration node.

[0158] This embodiment calculates a second similarity between the static attributes of the third candidate node and the original edge node, providing a supplementary decision based on similarity when no consensus is reached on the ranking. This effectively avoids the situation where there is no shared node index number and still allows the node to be migrated that is closest to the original edge node to be found.

[0159] The method further comprises:

[0160] S8. Migrate the computing task to the optimal migration node.

[0161] Specifically, task migration can be implemented using existing edge computing task migration technologies, such as saving and restoring task context, encapsulating and transmitting running status, and reconstructing the environment of the target node, so as to smoothly migrate computing tasks from the original node to the optimal migration node, thereby ensuring service continuity of task execution.

[0162] In this embodiment, by aligning the distance node sequence and the quality node sequence based on descending numbering, a node pair sequence is constructed, and a binary node pair with a shared node index number is sequentially searched to see whether there is one, thereby realizing a dual-dimensional fusion judgment of the operating status and the operating quality. When a shared node exists, it can be directly determined as the optimal migration node. If it is not hit, it is further screened through the static similarity of the third candidate node to ensure that the optimal migration node is accurately determined during task migration.

[0163] In summary, the present invention proposes a method for determining the optimal migration node based on double sequence alignment. Specifically, first, based on the first similarity between the static attribute vector of the original edge node and the static attribute vectors of several first candidate nodes, Q second candidate nodes whose static attributes are closest to the original edge node are screened out through static feature clustering and cluster center proximity matching. Then, the parameter regression model is used to predict the dynamic operating parameters of these Q second candidate nodes under the current computing task to calculate their dynamic operating quality and the Euclidean distance between them and the abnormal operating state. Subsequently, the dynamic operating quality of the second candidate node and the Euclidean distance to the abnormal node are numbered in descending order to construct two quality node sequences and a distance node sequence; then, based on the unified descending numbering, the node pair sequence is aligned to generate. On this basis, the optimal migration node is determined by sharing the node index number with the third candidate node.

[0164] Furthermore, if there is a second candidate node with a shared index number in the node pair, it means that the second candidate node performs best in both operating status and operating quality, and can be directly used as the optimal migration node;

[0165] If the shared node cannot be hit, the node determination path is supplemented by comparing the similarity of the static attributes of the end node corresponding to the largest descending number in the node sequence, ensuring that even if the sorting is inconsistent, the optimal migration target with the static feature closest to the original node can be selected.

[0166] The present invention gets rid of the dependence on a single sorting, avoids the conflict of "best quality" and "furthest status" nodes that cannot be uniformly judged, and effectively improves the accuracy and service continuity of edge computing tasks during task migration.

[0167] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired method (e.g., infrared, wireless, microwave, etc.).

[0168] The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., DVD ), or semiconductor media. The semiconductor media can be a solid-state drive.

[0169] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not implemented. In addition, the coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0170] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for dynamic migration of edge computing tasks supporting service continuity, characterized in that: include: S1. Anchor the original edge node where the computing task is running, and mark M first candidate nodes to which the computing task is to be migrated; S2. Assign unique node index numbers to the M first candidate nodes and the original edge node, and obtain the static attribute vectors corresponding to the node index numbers; S3. Screen Q second candidate nodes from the M first candidate nodes according to the static attribute vector corresponding to the node index number; S4. Concatenate the static attribute vector of the second candidate node and the standardized features of the computing task as joint inputs of a parameter regression model to predict several dynamic operating parameters of the computing task on the Q second candidate nodes; S5. Calculating the dynamic operation qualities of the Q second candidate nodes based on the predicted dynamic operation parameters, and calculating the Euclidean distances between the second candidate nodes and the original edge nodes; S6. Arrange the Q second candidate nodes in descending order according to the Euclidean distance and the dynamic operation quality to generate a distance node sequence and a quality node sequence; The distance node sequence and the quality node sequence are both arranged in descending order, and each second candidate node is assigned a descending number from the front to the back; S7, aligning the distance node sequence and the quality node sequence based on the descending numbering, generating a node pair sequence, and determining the optimal migration node in the node pair sequence; The node pair sequence includes Q binary node pairs arranged in descending order, and each binary node pair is paired with two second candidate nodes from the distance node sequence and the quality node sequence; The determining of the optimal migration node in the node pair sequence includes: S7-1. Determine whether there is a binary node pair sharing a node index number in the node pair sequence; S7-2. If there is a binary node pair that shares a node index number, mark the second candidate node of the binary node pair as the optimal migration node and terminate the search; otherwise, traverse the node pair sequence backward along the descending order of the binary node pairs; S7-3. If there is no binary node pair that shares a node index number, define the two second candidate nodes in the binary node pair corresponding to the largest descending number as the third candidate nodes; S7-4, determining the optimal migration node based on the third candidate node; Determining whether there is a binary node pair sharing a node index number in the node pair sequence includes: S7-1-1. In the node pair sequence, associate the node index numbers of the two second candidate nodes of each binary node pair; S7-1-2, search in descending order along the binary node pairs, and compare the node index numbers of the two second candidate nodes; S7-1-3. If the node index numbers of the two second candidate nodes are the same, the corresponding binary node pair is determined to be a binary node pair sharing the node index number; The searching for the optimal migration node based on the third candidate node includes: S7-4-1. Obtain the node index number of the third candidate node, and extract the static attribute vectors of the two third candidate nodes based on the node index number; S7-4-2. Calculate the second similarity between the static attribute vectors of the two third candidate nodes and the original edge node vector; S7-4-3. Compare the two second similarities, and determine the second candidate node with the larger second similarity as the optimal migration node; S8. Migrate the computing task to the optimal migration node.

2. A method for dynamic migration of edge computing tasks supporting service continuity according to claim 1, characterized in that: Marking the M first candidate nodes to which the computing task is to be migrated includes: S1-1. Obtaining the dynamic running quality of the computing task on the original edge node; S1-2. Determine whether the original edge node is a degenerate node based on the dynamic operation quality; S1-3. If it is a degenerate node, mark the M first candidate nodes to which the computing task is to be migrated; otherwise, continue to run the computing task on the original edge node.

3. A method for dynamic migration of edge computing tasks supporting service continuity according to claim 2, characterized in that: Obtaining the dynamic running quality of the computing task on the original edge node includes: S1-1-1. Slidingly collect several dynamic operating parameters of the computing task at the original edge node within a continuous time window; S1-1-2. Calculate the dynamic operation quality of the computing task within the time window based on several dynamic operation parameters of the original edge node.

4. A method for dynamic migration of edge computing tasks supporting service continuity according to claim 1, characterized in that: Assign unique node index numbers to the M first candidate nodes and the original edge nodes, including: S2-1, mark several edge nodes that have network connections with the original edge node; S2-2. Calculate the spatial distances between several edge nodes and the original edge node; S2-3, marking the edge node whose spatial distance is lower than the threshold as the first candidate node; S2-4. Assign unique node index numbers to the first candidate node and the original edge node according to the spatial distance between the first candidate node and the original edge node.

5. The method for dynamically migrating edge computing tasks supporting service continuity according to claim 1, characterized in that: According to the static attribute vector corresponding to the node index number, Q second candidate nodes are screened from the M first candidate nodes, including: S3-1. Perform K-nearest-neighbor clustering on the static attribute vectors of the M first candidate nodes to obtain K static clusters; S3-2, average pooling is performed on several static attribute vectors within the K static clusters to generate K cluster center vectors corresponding to the static clusters; S3-3, obtaining a static attribute vector of the original edge node, and calculating a first similarity between the static attribute vector and the K cluster center vectors; S3-4. Sort the K static clusters corresponding to the cluster center vectors according to the K first similarities to generate a cluster sequence; S3-5, starting from the static cluster corresponding to the maximum first similarity, intercepting the first Q similar clusters in the cluster sequence; S3-6, extract the cluster center vectors of the first Q similar clusters from the K cluster center vectors, and anchor the static attribute vector closest to the cluster center vector in each similar cluster; S3-7. Obtain the nearest static attribute vector and its node index number, and select Q second candidate nodes based on the node index number.

6. A method for dynamic migration of edge computing tasks supporting service continuity according to claim 1, characterized in that: Calculate the Euclidean distance between the second candidate node and the original edge node, including: S5-1. Mark the timestamp of when the original edge node is determined to be a degenerate node on the time axis; S5-2. Mark the dynamic operating parameters of the original edge node at the timestamp as abnormal operating parameters; S5-3. Normalize the features of several abnormal operation parameters of the original edge node and combine them into an abnormal operation vector of the original edge node; S5-4. Normalize the characteristics of several dynamic operation parameters of the Q second candidate nodes and concatenate them into dynamic operation vectors of the Q second candidate nodes; S5-5. Calculate Q Euclidean distances between the dynamic operation vector and the abnormal operation vector.

Images