A low-latency interactive intelligent network selection method based on a hybrid cloud architecture

Abstract

The application discloses a low-latency interactive intelligent network selection method based on a hybrid cloud architecture and relates to the technical field of cloud network fusion communication, which comprises the following steps: in the optimal path connection layer, the transmission state of an access network and the transmission state of a cloud node are combined to perform cooperative optimization, and a current competitive path set is formed; in the current competitive path set, shadow interactive messages which maintain the same window and sequence as current service messages are constructed, the shadow interactive messages are compared with real interactive messages in sequence, the shadow interactive confirmation quantity of each candidate cloud-edge transmission path is formed, and a target transmission path is selected according to the shadow interactive confirmation quantity; a service transmission channel is established along the target transmission path, parallel migration is performed between the current transmission path and the target transmission path, the real migration confirmation quantity is formed, and path replacement is completed when the real migration confirmation quantity is not worse than the shadow interactive confirmation quantity, thereby forming a new current transmission path. The application realizes cooperative control of candidate path screening and path replacement.

Description

Technical Field

[0001] This invention relates to the field of cloud-network converged communication technology, and in particular to a low-latency interactive intelligent network selection method based on a hybrid cloud architecture. Background Technology

[0002] In a hybrid cloud architecture, low-latency interactive services typically involve collaborative transmission across access-side network resources and cloud-side bearer resources. Related technologies need to balance operational requirements such as service requests, feedback, node capacity, path adjustment, and service continuity, while also meeting the transmission organization needs under conditions of multiple access points, multiple nodes, and multiple paths. Existing technologies generally select and adjust service paths based on link status, node response, and service bearer conditions to support continuous interactive services in cloud-network converged scenarios.

[0003] However, existing methods still have two limitations: First, they focus on directly screening candidate paths based on transmission status, without giving sufficient consideration to the continuity of paths when they enter current services, making it difficult to balance low latency requirements with smooth path switching. Second, once candidate paths are determined, there is often a lack of pre-verification and migration confirmation mechanisms that fit the service transmission rhythm, resulting in insufficient judgment of the target path's continuity capacity. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a low-latency interactive intelligent network selection method based on a hybrid cloud architecture, which solves the problems of difficulty in balancing path connectivity and low latency requirements during candidate path selection, as well as insufficient determination of target path connectivity.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a low-latency interactive intelligent network selection method based on a hybrid cloud architecture. The method includes: collecting access latency samples from the access network and cloud-side backhaul latency samples from cloud nodes, combining them to form candidate cloud-edge transmission paths, and comparing the path continuity relationships with existing transmission paths to form a candidate path record set containing access network transmission status, cloud node transmission status, and path continuity relationships; hierarchically sorting the candidate cloud-edge transmission paths according to the path continuity relationships between them and existing transmission paths, and performing collaborative optimization within the optimal path continuity layer by combining access network transmission status and cloud node transmission status to form a current competitive path set; constructing shadow interaction messages within the current competitive path set that maintain the same window and order as existing service messages, comparing the shadow interaction receipts with the real interaction receipts sequentially to form the shadow interaction confirmation quantity for each candidate cloud-edge transmission path, and selecting a target transmission path based on the shadow interaction confirmation quantity; establishing a service transmission channel along the target transmission path, performing parallel migration between the existing transmission path and the target path to form a real migration confirmation quantity, and completing path replacement when the real migration confirmation quantity is not inferior to the shadow interaction confirmation quantity to form a new existing transmission path.

[0007] As a preferred embodiment of the low-latency interactive intelligent network selection method based on a hybrid cloud architecture described in this invention, the method for collecting access latency samples of the access network and cloud-side backhaul latency samples of cloud nodes includes: during the transmission of service packets, recording the current transmission path for uplink transmission of service packets and downlink return of receipts as the current transmission path; initiating probe packet transmission and reception to the access network, recording the sending time and return packet arrival time of each probe packet, and recording the time length between the sending time and the return packet arrival time as the access latency sample; initiating service probe packet transmission and reception to cloud nodes in the cloud-side collaborative platform, recording the sending time and return packet arrival time of each service probe packet, and recording the time length between the sending time and the return packet arrival time as the cloud-side backhaul latency sample.

[0008] As a preferred embodiment of the low-latency interactive intelligent network selection method based on a hybrid cloud architecture described in this invention, the step of combining candidate cloud-edge transmission paths and comparing path continuity relationships with existing transmission paths includes: using the access network that has completed access latency sample collection as the service entry point, and sequentially sending service probe messages to the cloud nodes that have completed cloud-side backhaul latency sample collection; when a service probe message arrives at a cloud node after being sent from an access network and a service probe response packet is received, associating the access network that sent the service probe message with the cloud node that returned the service probe response packet, as a candidate cloud-edge transmission path; comparing the path continuity relationships between the candidate cloud-edge transmission path and the existing transmission path, and comparing based on service session mapping relationships, access network change relationships, and cloud node change relationships to form service session mapping states, access change relationships, and cloud node change relationships, and using the mapping continuation, mapping reconstruction, access continuation, access switching, cloud node continuation, and cloud node switching corresponding to the service session mapping states, access change relationships, and cloud node change relationships as path continuity relationships.

[0009] As a preferred embodiment of the low-latency interactive intelligent network selection method based on a hybrid cloud architecture described in this invention, the candidate path record set includes: statistically analyzing all access latency samples to form access latency values, access jitter values, and packet loss aggregation values, and using the access latency values, access jitter values, and packet loss aggregation values ​​as the access network transmission status; statistically analyzing all cloud-side backhaul latency samples of the same cloud node to form cloud-side backhaul latency values, and using the cloud-side backhaul latency values ​​as the cloud node transmission status; and associating and recording the candidate cloud-side transmission paths, access network transmission status, cloud node transmission status, and path connection relationships to form a candidate path record set.

[0010] As a preferred embodiment of the low-latency interactive intelligent network selection method based on hybrid cloud architecture described in this invention, the step of hierarchical sorting according to the path continuity relationship between candidate cloud-edge transmission paths and existing transmission paths includes: extracting candidate cloud-edge transmission paths and their associated records one by one from the candidate path record set; taking candidate cloud-edge transmission paths with service session mapping status of "mapping continued", access change relationship of "access continued", and cloud node change relationship of "cloud node continued" as existing retained paths; for candidate cloud-edge transmission paths other than existing retained paths, hierarchical recording is performed based on service session mapping status, access change relationship, and cloud node change relationship to form a path continuity layer; the path continuity layer is checked for its order, and the path continuity layer that is ranked first and contains candidate cloud-edge transmission paths is taken as the optimal path continuity layer.

[0011] As a preferred embodiment of the low-latency interactive intelligent network selection method based on a hybrid cloud architecture described in this invention, the step of performing collaborative optimization by combining the access network transmission status and the cloud node transmission status within the optimal path continuation layer includes: when the optimal path continuation layer has been formed, comparing the access network transmission status according to the access latency value, access jitter value, and packet loss aggregation value recorded by each candidate cloud-edge transmission path in the optimal path continuation layer to form an access priority order; comparing the cloud node transmission status according to the cloud-side backhaul latency value recorded by each candidate cloud-edge transmission path to form a cloud-side priority order; performing collaborative optimization on the candidate cloud-edge transmission paths in the optimal path continuation layer based on the access priority order and the cloud-side priority order; summarizing the candidate cloud-edge transmission paths retained in the optimal path continuation layer after collaborative optimization to form the current competitive path set; if the optimal path continuation layer has not been formed, then the currently retained path is taken as the only candidate cloud-edge transmission path in the current competitive path set.

[0012] As a preferred embodiment of the low-latency interactive intelligent network selection method based on hybrid cloud architecture described in this invention, the method for forming the shadow interaction confirmation quantity for each candidate cloud-edge transmission path includes: extracting each candidate cloud-edge transmission path from the current competitive path set and using the arrival time of the actual interaction receipt formed during the transmission of service messages in the current transmission path as a comparison benchmark; performing shadow interaction verification on the candidate cloud-edge transmission paths in the current competitive path set to form shadow interaction messages that are in the same window and order as the current service messages, and recording the arrival time of the shadow interaction receipt for each candidate cloud-edge transmission path; comparing the arrival time of the shadow interaction receipt in each candidate cloud-edge transmission path with the arrival time of the actual interaction receipt for the same message sequence number in the current transmission path to form a shadow receipt lead difference; forming a lead continuous segment length and a shadow interaction recovery value based on the shadow receipt lead difference, using the lead continuous segment length as the shadow acceptance base quantity, and using the shadow interaction recovery value to shrink the shadow acceptance base quantity to form the shadow interaction confirmation quantity.

[0013] As a preferred embodiment of the low-latency interactive intelligent network selection method based on hybrid cloud architecture described in this invention, the step of selecting the target transmission path according to the shadow interaction confirmation quantity includes: when the current competitive path set includes only one candidate cloud-edge transmission path, this candidate cloud-edge transmission path is taken as the target transmission path; when the current competitive path set includes multiple candidate cloud-edge transmission paths, they are sorted according to the shadow interaction confirmation quantity; when the shadow interaction confirmation quantities of two candidate cloud-edge transmission paths are the same, the leading continuous segment length and the shadow interaction recovery value are compared in turn; when the shadow interaction confirmation quantity, leading continuous segment length, and shadow interaction recovery value of two candidate cloud-edge transmission paths are all the same, the original sorting position of the two candidate cloud-edge transmission paths in the current competitive path set remains unchanged, forming a shadow interaction priority order, and the candidate cloud-edge transmission path ranked first in the shadow interaction priority order is taken as the target transmission path; the shadow interaction confirmation quantity, leading continuous segment length, and shadow interaction recovery value corresponding to the target transmission path are associated and recorded to form a target path record.

[0014] As a preferred embodiment of the low-latency interactive intelligent network selection method based on a hybrid cloud architecture described in this invention, the step of establishing a service transmission channel along the target transmission path and performing parallel migration between the current transmission path and the target transmission path includes: searching for the record of the target transmission path in the candidate path record set, and extracting the service session mapping status, access change relationship, and cloud node change relationship; when the target transmission path is inconsistent with the current transmission path, establishing a service transmission channel on the target transmission path according to the service session mapping status, access change relationship, and cloud node change relationship, and after the service transmission channel is established, synchronously sending the current service packets to the current transmission path and the target transmission path, performing parallel migration, and recording the arrival time of the migration receipt returned to the service-side device for each migration service packet.

[0015] As a preferred embodiment of the low-latency interactive intelligent network selection method based on a hybrid cloud architecture described in this invention, the step of forming a new current transmission path includes: comparing the arrival time of the migration receipt in the target transmission path with the arrival time of the actual interactive receipt for the same message sequence number in the current transmission path one by one to form a migration receipt lead difference; forming a migration lead continuous segment length and a migration recovery value based on the migration receipt lead difference, and using the migration lead continuous segment length as the migration acceptance basis quantity, and using the migration recovery value to shrink the migration acceptance basis quantity to form a true migration confirmation quantity; comparing the true migration confirmation quantity with the shadow interactive confirmation quantity in the target path record, and replacing the target transmission path with a new current transmission path when the true migration confirmation quantity is not inferior to the shadow interactive confirmation quantity.

[0016] The beneficial effects of this invention are as follows: By hierarchically sorting the path continuity relationships, the differences in the acceptance of candidate cloud-edge transmission paths when they enter the current service transmission are constrained. The path continuity layer that is ranked first and contains candidate cloud-edge transmission paths is limited to the candidate range. Then, within the candidate range, collaborative optimization is performed by combining the access network transmission status and the cloud node transmission status. This achieves phased screening of the acceptance relationship and transmission status of candidate cloud-edge transmission paths, so that the current set of competing paths can centrally represent candidate cloud-edge transmission paths with smaller acceptance differences and better transmission status. By connecting the shadow interaction confirmation quantity and the actual migration confirmation quantity, the continuous confirmation of the acceptance performance of the target transmission path is achieved, so that the formation of new current transmission paths has more sufficient judgment basis. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a low-latency interactive intelligent network selection method based on a hybrid cloud architecture.

[0019] Figure 2 This is a schematic diagram of the current set of competing paths.

[0020] Figure 3 This is a schematic diagram of the target transmission path.

[0021] Figure 4 This is a schematic diagram of the new existing transmission path.

[0022] Figure 5 This is a graph validating the joint screening effect of the current set of competing paths.

[0023] Figure 6 Verification diagram for connecting shadow interaction confirmation volume with actual migration confirmation volume. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0027] Reference Figures 1-6 This is one embodiment of the present invention, which provides a low-latency interactive intelligent network selection method based on a hybrid cloud architecture, including the following steps: In this invention, network selection refers to the joint selection of candidate cloud-edge transmission paths formed by the combination of access networks and cloud nodes in cloud-network converged service scenarios.

[0028] S1. Collect access latency samples from the access network and cloud-side backhaul latency samples from the cloud nodes, combine them to form candidate cloud-side transmission paths, and compare the path connection relationships with the existing transmission paths to form a candidate path record set that includes the access network transmission status, cloud node transmission status, and path connection relationships.

[0029] Furthermore, during the transmission of service messages, the transmission path currently responsible for the uplink transmission of service messages and the downlink return of receipts is recorded as the current transmission path, and the current access network identifier and the current cloud node identifier in the current transmission path are recorded simultaneously.

[0030] The current transmission path includes a current access network and a current cloud node. The current access network refers to the access network currently used by the current transmission path, and the current cloud node refers to the cloud node currently used by the current transmission path. The access network refers to the network entry point in which the service-side equipment can establish a link connection and complete the sending and receiving of probe messages. The cloud node refers to the node in the cloud-side collaborative platform that can undertake service forwarding, service relay, or session mapping.

[0031] The current access network identifier and the current cloud node identifier are used as the benchmark information for comparing path connection relationships.

[0032] Furthermore, probe messages are sent and received to the access network, the sending time and return packet arrival time of each probe message are recorded, and the time length between the sending time and the return packet arrival time is recorded as the access delay sample.

[0033] Within the same acquisition period, access delay samples generated by each probe are recorded sequentially along a unified sampling rhythm. All access delay samples are arranged in ascending order of value, and the access delay sample in the middle position is taken as the access delay value to characterize the stable latency level of the access network under the current transmission state. The upper quartile and lower quartile values ​​are extracted from all access delay samples, and the difference between the upper quartile and lower quartile values ​​is taken as the access jitter value to characterize the access delay fluctuation amplitude. During the transmission and reception of probe messages, if there are two or more probe messages arranged in the order of transmission and neither of them has received a reply, the number of messages that have not received a reply is recorded as a continuous loss segment length. All continuous loss segment lengths are recorded for the same access network, and all continuous loss segment lengths are arranged in ascending order of value. The continuous loss segment length in the middle position is taken as the packet loss cluster value to characterize the concentration of continuous loss phenomena in the access network.

[0034] Access latency, access jitter, and packet loss aggregation are used as the access network transmission status.

[0035] Furthermore, the cloud nodes in the cloud-side collaborative platform initiate the sending and receiving of service probe messages, record the sending time and return packet arrival time of each service probe message, and record the time length between the sending time and the return packet arrival time as the cloud-side backhaul latency sample.

[0036] The cloud-side backhaul latency samples formed by each cloud node are recorded sequentially along a unified sampling rhythm. All cloud-side backhaul latency samples of the same cloud node are arranged in ascending order of value, and the cloud-side backhaul latency sample in the middle position is taken as the cloud-side backhaul latency value to characterize the stable backhaul level of the cloud node under the current transmission state.

[0037] The cloud-side backhaul latency value is used as the cloud node transmission status.

[0038] Furthermore, the access network that has completed the collection of access latency samples is used as the service entry point. Service probe messages are sent sequentially to the cloud nodes that have completed the collection of cloud-side backhaul latency samples, and the sending access network identifier, the arriving cloud node identifier, and the service probe return packet status of each service probe message are recorded.

[0039] When a service probe message is sent from an access network and arrives at a cloud node, and a service probe response packet is received, it indicates that the access network and the cloud node have the capability to transmit service messages. The access network that sent the service probe message and the cloud node that returned the service probe response packet are associated as a candidate cloud-edge transmission path. A candidate cloud-edge transmission path refers to the alternative service bearer path through which the service message enters the transmission process from the access network and returns a response packet via the cloud node.

[0040] Furthermore, the candidate cloud-edge transmission paths are compared with the existing transmission paths to determine their path continuity.

[0041] Path connection relationship comparison refers to comparing the acceptance methods of candidate cloud-edge transmission paths when they are cut into the current service transmission, based on three aspects: service session mapping relationship, access network change relationship, and cloud node change relationship.

[0042] The existing service session mapping relationship carried by the cloud node is compared with the service session mapping relationship carried by the cloud node in the candidate cloud-edge transmission path. When the service session is switched to a cloud node in the candidate cloud-edge transmission path and the original service session mapping relationship is maintained, the current service session mapping state is used as the mapping to be continued. When the service session is switched to a cloud node in the candidate cloud-edge transmission path and the original service session mapping relationship is no longer maintained, and the service session mapping relationship is rebuilt, the current service session mapping state is used as the mapping to be rebuilt.

[0043] The access network identifier in the candidate cloud-edge transmission path is compared with the current access network identifier. When the access network in the candidate cloud-edge transmission path is the same as the current access network, the current access change relationship is used as the access continuation; when the access network in the candidate cloud-edge transmission path is different from the current access network, the current access change relationship is used as the access switching.

[0044] The cloud node identifiers in the candidate cloud-edge transmission path are compared with the current cloud node identifiers. When the cloud node in the candidate cloud-edge transmission path is consistent with the current cloud node, the current cloud node change relationship is used as the cloud node to be used. When the cloud node in the candidate cloud-edge transmission path is inconsistent with the current cloud node, the current cloud node change relationship is used as the cloud node to be switched.

[0045] The relationships between mapping reuse, mapping reconstruction, access reuse, access switching, cloud node reuse, and cloud node switching are defined as path continuation relationships.

[0046] Furthermore, candidate cloud-edge transmission paths, access latency values, access jitter values, packet loss aggregation values, cloud-side backhaul latency values, service session mapping status, access change relationships, and cloud node change relationships are associated and recorded to form a candidate path record set. The path records formed by the current access network and the current cloud nodes are also included in the candidate path record set.

[0047] The candidate path record set is used to collect the access network transmission status, cloud node transmission status, and path connection relationship comparison content of each candidate cloud-edge transmission path.

[0048] S2. Sort the candidate cloud-edge transmission paths hierarchically according to the path connection relationship between them and the current transmission paths. In the optimal path connection layer, perform collaborative optimization by combining the access network transmission status and the cloud node transmission status to form the current competitive path set.

[0049] Furthermore, candidate cloud-edge transmission paths and their associated records are extracted one by one from the candidate path record set. The service session mapping status, access change relationship and cloud node change relationship are used for path connection layer division to reflect the ease of acceptance when candidate cloud-edge transmission paths cut into existing transmission paths. The access network transmission status and cloud node transmission status are used for collaborative optimization within the optimal path connection layer.

[0050] Furthermore, candidate cloud-edge transmission paths with the service session mapping status set to "mapping continued", the access change relationship set to "access continued", and the cloud node change relationship set to "cloud node continued" will be retained as current reserved paths.

[0051] The existing reserved path is used to maintain the baseline position for current service transmissions.

[0052] Apart from the existing reserved paths, the remaining candidate cloud-edge transmission paths are sorted hierarchically according to their path connection relationships.

[0053] Furthermore, based on the service session mapping status, access change relationship, and cloud node change relationship, candidate cloud-edge transmission paths other than the currently reserved paths are recorded in layers to form a path connection layer. The path connection layer is a path grouping after the candidate cloud-edge transmission paths are layered according to the difficulty of acceptance based on the service session mapping status, access change relationship, and cloud node change relationship. The earlier the path connection layer is, the smaller the acceptance change caused when the candidate cloud-edge transmission path is cut into the current service transmission.

[0054] Candidate cloud-edge transmission paths whose service session mapping status is "mapping continued" and whose access change relationship and cloud node change relationship both show a switch are assigned to the first path continuation layer; candidate cloud-edge transmission paths whose service session mapping status is "mapping continued" and whose access change relationship and cloud node change relationship both show a switch are assigned to the second path continuation layer; candidate cloud-edge transmission paths whose service session mapping status is "mapping reconstructed" and whose access change relationship and cloud node change relationship both show a switch are assigned to the third path continuation layer; candidate cloud-edge transmission paths whose service session mapping status is "mapping reconstructed" and whose access change relationship and cloud node change relationship both show a switch are assigned to the fourth path continuation layer; through this processing, all candidate cloud-edge transmission paths in the candidate path record set are transformed into path continuation layers arranged according to the ease of connection.

[0055] The difficulty of connecting the first path continuation layer is lower than that of the second path continuation layer, the difficulty of connecting the second path continuation layer is lower than that of the third path continuation layer, and the difficulty of connecting the third path continuation layer is lower than that of the fourth path continuation layer.

[0056] Furthermore, after the path connection layer is formed, the distribution of candidate cloud-edge transmission paths in the first path connection layer, the second path connection layer, the third path connection layer and the fourth path connection layer are checked in turn, and the path connection layer that is ranked first and contains candidate cloud-edge transmission paths is selected as the optimal path connection layer.

[0057] When the first path continuation layer contains candidate cloud-edge transmission paths, the first path continuation layer is designated as the optimal path continuation layer; when the first path continuation layer does not contain candidate cloud-edge transmission paths but the second path continuation layer does, the second path continuation layer is designated as the optimal path continuation layer; when neither the first nor the second path continuation layer contains candidate cloud-edge transmission paths but the third path continuation layer does, the third path continuation layer is designated as the optimal path continuation layer; when none of the first, second, or third path continuation layers contain candidate cloud-edge transmission paths but the fourth path continuation layer does, the fourth path continuation layer is designated as the optimal path continuation layer.

[0058] If none of the first path continuation layer, the second path continuation layer, the third path continuation layer, and the fourth path continuation layer contain candidate cloud-edge transmission paths, then an optimal path continuation layer will not be formed, and the existing reserved path will be retained.

[0059] The optimal path continuation layer is used to limit the candidate range of the current competitive path set. When the path continuation layer ranked higher already contains candidate cloud-edge transmission paths, the candidate cloud-edge transmission paths in the path continuation layer ranked lower will not enter the current competitive path set. Through this process, the changes in acceptance caused by the candidate cloud-edge transmission path switching into the current service transmission are first constrained, and then subsequent comparisons are performed within the candidate range with smaller changes in acceptance.

[0060] Furthermore, once the optimal path connection layer has been formed, the access network transmission status comparison is performed on the candidate cloud-edge transmission paths in the optimal path connection layer according to the access latency value, access jitter value, and packet loss aggregation value recorded for each candidate cloud-edge transmission path. The access network transmission status comparison uses the access latency value as the first comparison basis, the access jitter value as the second comparison basis, and the packet loss aggregation value as the third comparison basis.

[0061] When the access latency of one candidate cloud-edge transmission path is less than that of another, the candidate cloud-edge transmission path with the smaller access latency is ranked first. When the access latency of two candidate cloud-edge transmission paths is the same, the access jitter value is compared. If the access jitter value of one candidate cloud-edge transmission path is less than that of another, the candidate cloud-edge transmission path with the smaller access jitter value is ranked first. When both the access latency and access jitter values ​​of two candidate cloud-edge transmission paths are the same, the packet loss aggregation value is compared. If the packet loss aggregation value of one candidate cloud-edge transmission path is less than that of another, the candidate cloud-edge transmission path with the smaller packet loss aggregation value is ranked first. When the access latency, access jitter, and packet loss aggregation values ​​of two candidate cloud-edge transmission paths are all the same, the two candidate cloud-edge transmission paths are recorded as having the same access priority order. Through the process of comparing access network transmission states, the access priority order of each candidate cloud-edge transmission path within the optimal path continuation layer is formed.

[0062] Furthermore, when the optimal path continuation layer has been formed and the access priority order has been formed, the cloud node transmission status comparison is performed on the candidate cloud edge transmission paths in the optimal path continuation layer according to the cloud side backhaul latency value recorded by each candidate cloud edge transmission path.

[0063] When the cloud-side backhaul latency of one candidate cloud-side transmission path is less than that of another candidate cloud-side transmission path, the candidate cloud-side transmission path with the smaller cloud-side backhaul latency is ranked first. When the cloud-side backhaul latency of two candidate cloud-side transmission paths is the same, the order of the two candidate cloud-side transmission paths in the access priority remains unchanged. After the cloud node transmission status comparison process, the cloud-side priority order of each candidate cloud-side transmission path in the optimal path connection layer is formed.

[0064] Furthermore, when the optimal path continuation layer has been formed and both the access priority order and the cloud-side priority order have been formed, collaborative optimization is performed on the candidate cloud-side transmission paths in the optimal path continuation layer.

[0065] For any two candidate cloud-edge transmission paths in the optimal path continuation layer, perform pairwise comparisons. If a candidate cloud-edge transmission path is ahead of another candidate cloud-edge transmission path in the access priority order and also ahead of another candidate cloud-edge transmission path in the cloud-side priority order, then the former candidate cloud-edge transmission path is designated as the preferred path, and the latter candidate cloud-edge transmission path is designated as the inferior path. If a candidate cloud-edge transmission path is ahead of another candidate cloud-edge transmission path in the access priority order but in the same order as another candidate cloud-edge transmission path in the cloud-side priority order, then the former candidate cloud-edge transmission path is designated as the preferred path, and the latter candidate cloud-edge transmission path is designated as the inferior path. The path is designated as the inferior path. When a candidate cloud-edge transmission path is in the same order as another candidate cloud-edge transmission path in the access priority order, but is ahead of the other candidate cloud-edge transmission path in the cloud-side priority order, the former candidate cloud-edge transmission path is designated as the superior path, and the latter candidate cloud-edge transmission path is designated as the inferior path. When two candidate cloud-edge transmission paths are in the same order in both the access priority order and the cloud-side priority order, both candidate cloud-edge transmission paths are retained. After all pairwise comparisons are completed, the candidate cloud-edge transmission path designated as the inferior path is removed from the optimal path continuation layer, and the candidate cloud-edge transmission paths not designated as inferior paths are retained in the optimal path continuation layer.

[0066] The candidate cloud-edge transmission paths that remain in the optimal path continuation layer after collaborative optimization are considered as candidate cloud-edge transmission paths with advantages in carrying current service transmission conditions.

[0067] Furthermore, the candidate cloud-edge transmission paths retained in the optimal path continuation layer after collaborative optimization are aggregated to form the current competitive path set.

[0068] If the optimal path continuation layer has been formed, and there are candidate cloud-edge transmission paths retained after collaborative optimization in the optimal path continuation layer, then the candidate cloud-edge transmission paths retained after collaborative optimization will be used as the current competitive path set; if the optimal path continuation layer has not been formed, then the currently retained path will be used as the only candidate cloud-edge transmission path in the current competitive path set.

[0069] It should be noted that the comparison of access network transmission status, cloud node transmission status, and collaborative optimization are only performed in the optimal path continuation layer. This is used to further filter candidate cloud-edge transmission paths with better transmission status among those with less change. The current set of competing paths formed in this way is not obtained by directly sorting all candidate cloud-edge transmission paths by latency, but is formed within the candidate range limited by the path continuation relationship.

[0070] The candidate cloud-edge transmission paths in the current competitive path set retain the access network transmission status, cloud node transmission status, and path connection relationship to characterize the range of paths with the greatest carrying capacity under the current service transmission conditions.

[0071] S3. Construct shadow interaction messages that are in the same window and order as the current business messages within the current competitive path set. Compare the shadow interaction receipts with the real interaction receipts by sequence number to form the shadow interaction confirmation quantity of each candidate cloud-edge transmission path, and select the target transmission path according to the shadow interaction confirmation quantity.

[0072] Furthermore, candidate cloud-edge transmission paths in the current competitive path set are extracted one by one, and the arrival time of the real interactive receipt formed during the transmission of service messages in the current transmission path is used as the comparison benchmark. The arrival time of the real interactive receipt indicates the arrival time of the real interactive receipt returned to the service-side device after the transmission of each service message in the current transmission path is completed. It is used to characterize the actual receipt return time of the current transmission path at each message sequence number position.

[0073] Perform shadow interaction verification on the candidate cloud-edge transmission paths in the current competitive path set, and record the arrival time of the shadow interaction receipt for each candidate cloud-edge transmission path. The arrival time of the shadow interaction receipt indicates the arrival time of the shadow interaction receipt after the shadow interaction message is transmitted along the candidate cloud-edge transmission path and returned to the service-side device. It is used to characterize the receipt return time of the candidate cloud-edge transmission path at each message sequence number position. The arrival time of the shadow interaction receipt is used to compare the sequence number with the arrival time of the real interaction receipt.

[0074] Shadow interaction verification refers to copying the sending time, sequence number, and length of the current service message to the candidate cloud-edge transmission path without changing the current service transmission path, forming shadow interaction messages with the same window and sequence. Same window and sequence means that the shadow interaction message and the current service message are located in the same sending time window and maintain the same message sequence number and message length.

[0075] To ensure that shadow interaction messages are used only for path verification, a shadow verification identifier is written into the shadow interaction message. After the cloud node recognizes the shadow verification identifier, it only returns a shadow interaction receipt and does not send the shadow interaction message into the actual business execution process.

[0076] Furthermore, the arrival time of the shadow interaction receipt in each candidate cloud-edge transmission path is compared one by one with the arrival time of the actual interaction receipt for the same message sequence number in the current transmission path to form the shadow receipt lead difference, which is expressed as: ; in, Indicates the first The candidate cloud-edge transmission path is in the first The leading difference of shadow receipt on each message sequence number; Indicates the first The candidate cloud-edge transmission path is in the first The arrival time of the shadow interactive receipt on the message sequence number; Indicates the current transmission path is at the 1st rank. The actual arrival time of the interactive receipt on the message sequence number; Indicates the candidate cloud-edge transmission path number in the current competitive path set; This indicates the sequence number of the message within the same transmission time window.

[0077] The shadow acknowledgment lead difference is used to characterize the arrival order of shadow interaction acknowledgments in the candidate cloud-edge transmission path relative to the actual interaction acknowledgments in the current transmission path.

[0078] When the lead difference of the shadow receipt is less than zero, it means that the shadow interaction receipt arrives before the real interaction receipt; when the lead difference of the shadow receipt is greater than zero, it means that the shadow interaction receipt arrives after the real interaction receipt; when the lead difference of the shadow receipt is equal to zero, it means that the shadow interaction receipt and the real interaction receipt arrive at the same time.

[0079] Furthermore, the lead difference of all shadow receipts formed by the same candidate cloud-edge transmission path is checked sequentially from smallest to largest along the message sequence number. Records with a lead difference of less than zero and consecutive message sequence numbers are grouped into a lead segment, and the number of message sequence numbers in each lead segment is counted. The maximum number of message sequence numbers in all lead segments is taken as the length of the lead continuous segment.

[0080] If the leading difference of the shadow receipts is not less than zero, the length of the leading continuous segment is recorded as zero.

[0081] The length of the leading continuum segment is used to reflect the ability of a candidate cloud-edge transmission path to continuously lead the existing transmission path within the same transmission time window; the larger the length of the leading continuum segment, the stronger the candidate cloud-edge transmission path's ability to continuously lead.

[0082] Furthermore, when a certain leading segment ends, the shadow acknowledgment leading difference becomes not less than zero, and then becomes less than zero again in a later message sequence number position, the message sequence number difference between the first non-leading message sequence number after the end of the previous leading segment and the first leading message sequence number at the beginning of the next leading segment is used as a shadow interactive recovery span.

[0083] All shadow interaction recovery spans are aggregated along the same candidate cloud-edge transmission path, and all shadow interaction recovery spans are arranged in ascending order of value. The shadow interaction recovery span in the middle position is taken as the shadow interaction recovery value.

[0084] If a leading segment exists but no shadow interaction recovery span is formed, the shadow interaction recovery value is recorded as zero; if no leading segment is formed, the number of message sequence numbers participating in the comparison is used as the shadow interaction recovery value; the number of message sequence numbers participating in the comparison is at least one.

[0085] The shadow interaction recovery value is used to reflect how quickly a candidate cloud-edge transmission path recovers its lead after falling behind in a segment; the smaller the shadow interaction recovery value, the stronger the candidate cloud-edge transmission path's ability to recover its lead after a short period of lagging behind.

[0086] Furthermore, the length of the leading continuous segment is used as the basic quantity for shadow acceptance, and the shadow interaction recovery value is used to shrink the basic quantity for shadow acceptance to form the shadow interaction confirmation quantity, expressed as: ; in, Indicates the first Shadow interaction confirmation quantity of candidate cloud-edge transmission paths; Indicates the first The length of the leading contiguous segment of each candidate cloud-edge transmission path; Indicates the first Shadow interaction recovery value of candidate cloud-edge transmission paths.

[0087] The shadow interaction confirmation quantity is used to characterize the capacity of candidate cloud-edge transmission paths under the combined effect of continuous leading capability and recovery leading capability; the larger the shadow interaction confirmation quantity, the stronger the candidate cloud-edge transmission path's ability to undertake current service transmission.

[0088] The value of the shadow interaction confirmation quantity is between zero and one; when the length of the leading continuous segment increases while the shadow interaction recovery value remains unchanged, the shadow interaction confirmation quantity increases accordingly; when the shadow interaction recovery value increases while the length of the leading continuous segment remains unchanged, the shadow interaction confirmation quantity decreases accordingly.

[0089] Furthermore, when the current competitive path set includes only one candidate cloud-edge transmission path, the shadow interaction confirmation quantity formed by this candidate cloud-edge transmission path is used as the sole ranking criterion, and this candidate cloud-edge transmission path is used as the target transmission path.

[0090] Furthermore, when the current competitive path set includes multiple candidate cloud-edge transmission paths, the candidate cloud-edge transmission paths in the current competitive path set are sorted according to the number of shadow interaction confirmations.

[0091] When the shadow interaction acknowledgment value of one candidate cloud-edge transmission path is greater than that of another, the candidate cloud-edge transmission path with the larger shadow interaction acknowledgment value is ranked higher. When the shadow interaction acknowledgment values ​​of two candidate cloud-edge transmission paths are the same, the leading continuous segment length is compared. If the leading continuous segment length of one candidate cloud-edge transmission path is greater than that of another, the candidate cloud-edge transmission path with the larger leading continuous segment length is ranked higher. When both the shadow interaction acknowledgment value and the leading continuous segment length of two candidate cloud-edge transmission paths are the same, the shadow interaction recovery value is compared. If the shadow interaction recovery value of one candidate cloud-edge transmission path is less than that of the other, the candidate cloud-edge transmission path is ranked higher. When recovering the shadow interaction value of a side transmission path, the candidate cloud-side transmission path with the smaller shadow interaction recovery value is ranked first. When the shadow interaction acknowledgment, leading continuous segment length, and shadow interaction recovery value of two candidate cloud-side transmission paths are all the same, the original ranking of the two candidate cloud-side transmission paths in the current competitive path set remains unchanged. The original ranking is formed with access priority as the primary order and cloud-side priority as the secondary order. When two candidate cloud-side transmission paths are in the same order in both access priority and cloud-side priority, the order in which the two candidate cloud-side transmission paths entered the current competitive path set remains unchanged. After this processing, the shadow interaction priority order in the current competitive path set is formed.

[0092] The candidate cloud-edge transmission path ranked first in the shadow interaction priority order is taken as the target transmission path, and the shadow interaction confirmation quantity, leading continuous segment length and shadow interaction recovery value corresponding to the target transmission path are associated and recorded to form the target path record.

[0093] The target path record is used to compile the sorting criteria for the target transmission path in the peer-to-peer shadow interaction verification.

[0094] S4. Establish a service transmission channel along the target transmission path, perform parallel migration between the current transmission path and the target transmission path to form a real migration confirmation quantity, and complete the path replacement when the real migration confirmation quantity is not inferior to the shadow interaction confirmation quantity to form a new current transmission path.

[0095] Furthermore, the target transmission path is extracted from the target path record, and the record containing the target transmission path is searched in the candidate path record set to extract the business session mapping status, access change relationship, and cloud node change relationship.

[0096] When the target transmission path is the same as the current transmission path, the current transmission path will be retained as the new current transmission path.

[0097] When the target transmission path differs from the existing transmission path, a service transmission channel is established on the target transmission path according to the service session mapping status, access change relationship, and cloud node change relationship. When the service session mapping status is "mapping continued," the service session mapping relationship carried by the existing cloud node is extended to the cloud node in the target transmission path, serving as the service session mapping relationship for the target transmission path. When the service session mapping status is "mapping reconstruction," a new service session mapping relationship is established on the cloud node in the target transmission path, and this newly established relationship serves as the service session mapping relationship for the target transmission path. When the access change relationship is "access continued," the existing access network remains unchanged. When the access change relationship is "access switching," the existing access network carries the current service packets, and the access network in the target transmission path carries the migration service packets. When the cloud node change relationship is "cloud node continued," the existing cloud node remains unchanged. When the cloud node change relationship is "cloud node switching," the existing cloud node carries the current service packets, and the cloud node in the target transmission path carries the migration service packets. Through this process, a service transmission channel capable of participating in parallel migration is formed on the target transmission path.

[0098] Furthermore, when the target transmission path is inconsistent with the current transmission path, and after the service transmission channel on the target transmission path is established, the current service messages are synchronously sent to both the current transmission path and the target transmission path to perform parallel migration.

[0099] Service messages entering the current transmission path continue to complete the current actual service transmission, while service messages entering the target transmission path participate in the same batch of service processing as migrated service messages.

[0100] Migrating service messages and existing service messages are located within the same sending time window, and maintain consistency in sending time, message sequence number, and message length.

[0101] To ensure that the target transmission path undergoes migration verification under real business conditions, no shadow verification identifier is written in the migration service message. After receiving the migration service message, the cloud node in the target transmission path enters the actual business processing link and generates the corresponding business processing result. It only returns the migration receipt to the business-side device and keeps the corresponding business processing result inside the cloud node in the target transmission path, without submitting the final business result to the outside world.

[0102] Record the arrival time of the migration receipt returned to the service-side device for each migration service message; the arrival time of the migration receipt is used to compare the sequence number with the arrival time of the actual interactive receipt in the current transmission path.

[0103] Furthermore, the arrival time of the migration receipt in the target transmission path is compared one by one with the arrival time of the actual interactive receipt for the same message sequence number in the current transmission path to form the migration receipt lead difference, which is expressed as: ; in, Indicates the first The migration receipt lead difference on each message sequence number; Indicates the target transmission path at the th The arrival time of the migration receipt on the message sequence number.

[0104] The migration acknowledgment lead difference is used to characterize the arrival order of migration acknowledgments in the target transmission path relative to the actual interactive acknowledgments in the current transmission path.

[0105] When the leading difference of the migration receipt is less than zero, it means that the migration receipt arrived before the actual interaction receipt; when the leading difference of the migration receipt is greater than zero, it means that the migration receipt arrived after the actual interaction receipt; when the leading difference of the migration receipt is equal to zero, it means that the migration receipt and the actual interaction receipt arrived at the same time.

[0106] Furthermore, the lead difference of all migration receipts formed by the target transmission path is checked one by one in ascending order of message sequence number.

[0107] During the process of checking all migration acknowledgment lead differences formed by the target transmission path, when the migration acknowledgment lead difference for a certain message sequence number is less than zero, this message sequence number is taken as the starting message sequence number of a migration lead segment; when the subsequent message sequence numbers are incremented by one relative to the previous message sequence number, and the corresponding migration acknowledgment lead difference is continuously less than zero, the subsequent message sequence numbers are continued to be merged into the same migration lead segment; when the subsequent message sequence numbers do not satisfy the condition of incrementing by one, or the corresponding migration acknowledgment lead difference is no longer less than zero, the statistics of the current migration lead segment are terminated.

[0108] For each migration-leading segment, the sequence number span between the segment's ending message number and the segment's starting message number is incremented by one to determine the current migration-leading segment's sequence number. The maximum sequence number among all migration-leading segments is taken as the migration-leading contiguous segment length. If no migration-leading segment is formed, the migration-leading contiguous segment length is recorded as zero.

[0109] The migration lead length is used to reflect the ability of the target transmission path to maintain a lead over the existing transmission path during parallel migration; the larger the migration lead length, the stronger the ability of the target transmission path to maintain a lead during parallel migration.

[0110] Furthermore, when a certain migration lead segment ends, the migration receipt lead difference becomes not less than zero, and then becomes less than zero again in a later message sequence number position, the message sequence number difference between the first non-leading message sequence number after the end of the previous migration lead segment and the first leading message sequence number at the beginning of the next migration lead segment is taken as a migration recovery span.

[0111] Summarize all migration recovery spans along the target transmission path, arrange all migration recovery spans in ascending order of value, and take the migration recovery span in the middle position as the migration recovery value.

[0112] If a migration leader segment exists but no migration recovery span is formed, the migration recovery value is recorded as zero; if no migration leader segment is formed, the number of message sequence numbers participating in the comparison is used as the migration recovery value; the number of message sequence numbers participating in the comparison is at least one.

[0113] The migration recovery value reflects how quickly the target transmission path recovers its lead after falling behind during parallel migration; the smaller the migration recovery value, the stronger the target transmission path's ability to recover its lead after a short period of lagging behind.

[0114] Furthermore, the length of the leading continuous segment of migration is used as the basis for migration acceptance, and the migration recovery value is used to shrink the basis for migration acceptance to form the actual migration confirmation quantity, expressed as: ; in, This represents the actual migration confirmation amount of the target transmission path; Indicates the length of the leading contiguous segment of the target transmission path; This represents the migration recovery value of the target transmission path.

[0115] The actual migration confirmation quantity is used to characterize the carrying capacity of the target transmission path under the combined effect of its sustained leading capability and recovery leading capability during parallel migration; the larger the actual migration confirmation quantity, the stronger the carrying capacity of the target transmission path during the migration process.

[0116] The true migration confirmation value is between zero and one. When the length of the migration leading segment increases while the migration recovery value remains unchanged, the true migration confirmation value increases accordingly. When the migration recovery value increases while the length of the migration leading segment remains unchanged, the true migration confirmation value decreases accordingly.

[0117] By using the migration lead length as the basis for migration acceptance and shrinking the migration acceptance basis using the migration recovery value, a joint constraint on the continuous lead capability and the degree of recovery lag is achieved. When the target transmission path forms a long migration lead length during parallel migration, the actual migration confirmation quantity can remain at a high level. When the target transmission path forms a large migration recovery value during parallel migration, the actual migration confirmation quantity will converge accordingly, thus making the actual migration confirmation quantity more consistent with the actual migration acceptance level of the target transmission path.

[0118] Furthermore, the actual migration confirmations are compared with the shadow interaction confirmations in the target path record.

[0119] Shadow interaction confirmation reflects the verification performance of the target transmission path without entering the actual business execution process, while real migration confirmation reflects the acceptance performance of the target transmission path after entering the actual business execution process during parallel migration.

[0120] The shadow interaction confirmation quantity serves as the baseline for the target transmission path before entering the parallel migration verification.

[0121] When the actual migration confirmation quantity is not worse than the shadow interaction confirmation quantity, it indicates that the target transmission path has maintained the acceptance level of the shadow interaction verification phase under real business conditions. At this time, the target transmission path is replaced with the new existing transmission path. When the actual migration confirmation quantity is worse than the shadow interaction confirmation quantity, it indicates that the target transmission path has experienced a decrease in acceptance during the actual migration process. At this time, the original existing transmission path is kept unchanged.

[0122] When the target transmission path is replaced by the new existing transmission path, the writing of service messages to the original existing transmission path is stopped, and the access network identifier and cloud node identifier in the new existing transmission path are updated to the current access network identifier and current cloud node identifier.

[0123] Furthermore, the new existing transmission path, actual migration confirmation quantity, migration leading continuous segment length, and migration recovery value are associated and recorded to form a migration path record.

[0124] The migration path record retains both the migration behavior during path replacement and the current path information after path replacement, so that the new current transmission path not only completes the service bearer location update, but also retains the actual migration process on which this replacement is based.

[0125] The migration path record is used to collect the migration basis and path replacement information of the target transmission path during the parallel migration process.

[0126] It should also be noted that, to verify the impact of hierarchical sorting of path continuity relationships and the collaborative optimization of access network transmission status and cloud node transmission status on the formation of the current competitive path set, and to further verify the supporting role of the connection judgment between shadow interaction confirmation and actual migration confirmation in path replacement, comparative verification was conducted in a low-latency interactive simulation environment of a hybrid cloud architecture. During the simulation, the service message length, sending time window length, sampling time, service session size, and number of candidate cloud-edge transmission paths were kept constant near their normal average values. Only the fluctuation level of access latency samples, the distribution level of consecutive lost segments, and the level of cloud-side backhaul latency samples were adjusted in combination. Each combination condition formed a simulation scenario, and they were numbered sequentially according to the order of combined disturbance from weakest to strongest, forming... Figure 5 The simulation scenario number is used. For each simulation scenario, access latency samples, cloud-side backhaul latency samples, current transmission path records, candidate path record sets, shadow interaction receipt arrival times, real interaction receipt arrival times, and migration receipt arrival times are recorded uniformly. Based on this, shadow interaction confirmation quantities and real migration confirmation quantities are formed to evaluate the carrying capacity of candidate cloud-side transmission paths and the basis for path replacement.

[0127] Figure 5 The candidate path selection results under different simulation scenarios are summarized and plotted. For each simulation scenario, a candidate path record set is first formed according to the steps of this invention. Then, the path continuation relationship is sorted hierarchically according to the service session mapping status, access change relationship, and cloud node change relationship. In the path continuation layer that is at the top of the sort and contains candidate cloud-edge transmission paths, collaborative optimization is performed by combining access latency value, access jitter value, packet loss aggregation value, and cloud-side backhaul latency value to form the current competitive path set. Subsequently, shadow interaction verification is performed on the candidate cloud-edge transmission paths in the current competitive path set. The shadow interaction confirmation value of each candidate cloud-edge transmission path is calculated, and the candidate cloud-edge transmission path with the largest value is taken as the plotting value of this invention in the current simulation scenario. Figure 5 The phrase "only perform path continuation relationship hierarchical sorting" means that the path continuation relationship hierarchical sorting process is retained, the collaborative optimization of access network transmission status and cloud node transmission status is not performed, the shadow interaction confirmation quantity is directly calculated for the candidate cloud-edge transmission paths in the top-ranked path continuation layer, and the maximum value is taken as the drawing value. Figure 5 The phrase "sorting directly based only on access network transmission status and cloud node transmission status" means that the hierarchical sorting of path connection relationships is not performed. Instead, the path is sorted directly based on access latency, access jitter, packet loss aggregation, and cloud-side backhaul latency. The shadow interaction confirmation quantity is then calculated for the top-ranked candidate cloud-side transmission paths, and the maximum value is taken as the plotting value.

[0128] like Figure 5As shown, the horizontal axis represents the simulation scenario number, and the vertical axis represents the shadow interaction confirmation quantity of the optimal candidate cloud-edge transmission path in the current competitive path set. Figure 5 The solid blue line represents the method of this invention, the solid orange line represents hierarchical sorting based solely on path continuity, and the solid green line represents direct sorting based solely on access network transmission status and cloud node transmission status. As the simulation scenario number increases, all three data lines show an overall downward trend, indicating that the difficulty of accepting candidate cloud-edge transmission paths increases synchronously as disturbances on the access side and cloud side intensify. The data line for this invention's method is generally higher than the other two data lines and remains higher in later scenarios, indicating that this invention first hierarchically sorts based on path continuity, limiting the path continuity layer containing candidate cloud-edge transmission paths to the candidate range, and then performs collaborative optimization within the candidate range by combining access network transmission status and cloud node transmission status. The resulting current competitive path set can more effectively retain candidate cloud-edge transmission paths with smaller acceptance differences and better transmission status. This demonstrates that this invention does not directly select paths from all candidate cloud-edge transmission paths based solely on transmission status, but rather first determines the candidate range based on path continuity, and then performs collaborative optimization within the candidate range, thereby improving the support capability of the current competitive path set for the selection of subsequent target transmission paths.

[0129] Figure 6 The target path record and migration path record are summarized and drawn based on all simulation scenarios. For each simulation scenario, the shadow interaction confirmation quantity is first calculated for candidate cloud-edge transmission paths within the current competitive path set, and the target transmission path is selected according to the shadow interaction confirmation quantity to form a target path record; then, a service transmission channel is established along the target transmission path, and parallel migration is performed between the current transmission path and the target transmission path. The arrival time of the migration receipt is recorded, and the length of the leading continuous segment of migration, the migration recovery value, and the actual migration confirmation quantity are calculated to form a migration path record. Figure 6 Each scatter point corresponds to the pairing result of a target path record and a migration path record. The horizontal axis is the shadow interaction confirmation quantity in the target path record, and the vertical axis is the actual migration confirmation quantity in the migration path record. This is used to represent the correspondence between the acceptance performance of the same target transmission path in the shadow interaction verification stage and the parallel migration stage.

[0130] like Figure 6 As shown, the horizontal axis represents the number of shadow interaction confirmations, and the vertical axis represents the number of actual migration confirmations. Figure 6The blue scatter dots represent target transmission path records where path replacement has been completed, while the orange scatter dots represent target transmission path records where the original existing transmission path is maintained. The diagonal dashed line represents the reference line for determining whether the actual migration confirmation quantity equals the shadow interaction confirmation quantity. When a scatter dot is above or coincides with the reference line, it indicates that the actual migration confirmation quantity is not inferior to the shadow interaction confirmation quantity, and the corresponding target transmission path maintains the acceptance level of the shadow interaction verification phase under real business conditions, thus completing the path replacement. When a scatter dot is below the reference line, it indicates that the target transmission path experienced a decrease in acceptance during parallel migration, thus maintaining the original existing transmission path. Figure 6 The two types of scattered points are clearly distributed along both sides of the judgment reference line, indicating that a continuous judgment relationship has been established between the shadow interaction confirmation quantity and the actual migration confirmation quantity. This can provide more sufficient basis for path replacement, thereby improving the reliability of the target transmission path carrying capacity judgment.

[0131] In summary, this invention: by hierarchically sorting the path continuity relationships, it constrains the differences in the acceptance of candidate cloud-edge transmission paths when they are integrated into the current service transmission. It then limits the path continuity layer that is at the top of the sort and contains candidate cloud-edge transmission paths to the candidate range. Within this candidate range, it performs collaborative optimization by combining the access network transmission status and the cloud node transmission status. This achieves phased screening of the acceptance relationships and transmission status of candidate cloud-edge transmission paths, enabling the current set of competing paths to centrally represent candidate cloud-edge transmission paths with smaller acceptance differences and better transmission status. Furthermore, by connecting the shadow interaction confirmation quantity with the actual migration confirmation quantity, it achieves continuous confirmation of the acceptance performance of the target transmission path, providing a more sufficient basis for the formation of new current transmission paths.

[0132] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A low-latency interactive intelligent network selection method based on a hybrid cloud architecture, characterized in that, include: Collect access latency samples from the access network and cloud-side backhaul latency samples from cloud nodes, combine them to form candidate cloud-side transmission paths, and compare the path connection relationships with the existing transmission paths to form a candidate path record set that includes access network transmission status, cloud node transmission status, and path connection relationships. The candidate cloud-edge transmission paths are sorted hierarchically according to the path continuity relationship between them and the existing transmission paths. Within the optimal path continuity layer, collaborative optimization is performed by combining the access network transmission status and the cloud node transmission status to form the current set of competing paths. Within the current set of competing paths, construct shadow interaction messages that are in the same window and order as the existing business messages. Compare the shadow interaction receipts with the real interaction receipts by sequence number to form the shadow interaction confirmation quantity for each candidate cloud-edge transmission path. Select the target transmission path according to the shadow interaction confirmation quantity. Establish a service transmission channel along the target transmission path, perform parallel migration between the current transmission path and the target transmission path to form a real migration confirmation quantity, and complete the path replacement when the real migration confirmation quantity is not inferior to the shadow interaction confirmation quantity to form a new current transmission path.

2. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 1, characterized in that, The collection of access latency samples from the access network and cloud-side backhaul latency samples from cloud nodes includes recording the current transmission path for uplink transmission of service messages and downlink return of receipts during the transmission of service messages, which serves as the current transmission path. Initiate probe message transmission and reception to the access network, record the sending time and return packet arrival time of each probe message, and record the time length between the sending time and the return packet arrival time as the access delay sample; Initiate the sending and receiving of service probe messages to cloud nodes in the cloud-side collaborative platform, record the sending time and return packet arrival time of each service probe message, and record the time length between the sending time and the return packet arrival time as the cloud-side backhaul latency sample.

3. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 1, characterized in that, The combination forms a candidate cloud-edge transmission path, and the path connection relationship is compared with the existing transmission path. This includes taking the access network that has completed the collection of access latency samples as the service entry point and sending service probe messages to the cloud nodes that have completed the collection of cloud-side backhaul latency samples in sequence. When a service probe message is sent from an access network and arrives at a cloud node, and a service probe response packet is received, the access network that sent the service probe message and the cloud node that returned the service probe response packet are associated as a candidate cloud-edge transmission path. The candidate cloud-edge transmission paths are compared with the existing transmission paths to determine their path continuity. Based on the comparison of service session mapping relationships, access network change relationships, and cloud node change relationships, service session mapping status, access change relationships, and cloud node change relationships are formed. The mapping continuation, mapping reconstruction, access continuation, access switching, cloud node continuation, and cloud node switching corresponding to the service session mapping status, access change relationships, and cloud node change relationships are used as path continuity relationships.

4. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 2 or 3, characterized in that, The candidate path record set includes statistical analysis of all access delay samples to form access delay value, access jitter value, and packet loss cluster value, and uses the access delay value, access jitter value, and packet loss cluster value as the access network transmission status. Statistical analysis is performed on all cloud-side backhaul latency samples of the same cloud node to form a cloud-side backhaul latency value, and the cloud-side backhaul latency value is used as the transmission status of the cloud node. The candidate cloud-edge transmission paths, access network transmission status, cloud node transmission status, and path connection relationships are associated and recorded to form a candidate path record set.

5. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 1, characterized in that, The hierarchical sorting based on the path connection relationship between candidate cloud-edge transmission paths and current transmission paths includes extracting candidate cloud-edge transmission paths and their associated records one by one from the candidate path record set, and taking the candidate cloud-edge transmission paths with the service session mapping status of mapping continued, the access change relationship of access continued, and the cloud node change relationship of cloud node continued as the current reserved paths. For candidate cloud-edge transmission paths other than the existing reserved paths, a hierarchical recording is performed based on the service session mapping status, access change relationship and cloud node change relationship to form a path connection layer; The path continuation layers are checked for their order, and the path continuation layer that is ranked first and contains candidate cloud-edge transmission paths is selected as the optimal path continuation layer.

6. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 5, characterized in that, The step of performing collaborative optimization by combining access network transmission status and cloud node transmission status within the optimal path continuation layer includes, when the optimal path continuation layer has been formed, comparing the access network transmission status according to the access latency value, access jitter value and packet loss aggregation value recorded by each candidate cloud-edge transmission path in the optimal path continuation layer to form an access priority order. The cloud node transmission status is compared according to the cloud-side backhaul latency value recorded in each candidate cloud-side transmission path to form a cloud-side priority order. Based on access priority and cloud-side priority, collaborative optimization is performed on the candidate cloud-edge transmission paths in the optimal path continuation layer. The candidate cloud-edge transmission paths retained in the optimal path continuation layer after collaborative optimization are summarized to form the current competitive path set. If the optimal path continuation layer has not been formed, the existing reserved path will be used as the only candidate cloud-edge transmission path in the current set of competing paths.

7. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 1, characterized in that, The formation of the shadow interaction confirmation quantity for each candidate cloud-edge transmission path includes extracting each candidate cloud-edge transmission path in the current competitive path set and using the arrival time of the real interaction receipt formed by the current transmission path during the transmission of service messages as the comparison benchmark. Perform shadow interaction verification on the candidate cloud-edge transmission paths in the current competitive path set, form shadow interaction messages that are in the same window and order as the current service messages, and record the arrival time of the shadow interaction receipt of each candidate cloud-edge transmission path. The arrival time of the shadow interactive receipt in each candidate cloud-edge transmission path is compared with the arrival time of the real interactive receipt with the same message sequence number in the current transmission path to form the shadow receipt lead difference. The leading continuous segment length and shadow interaction recovery value are formed based on the leading difference of the shadow receipt. The leading continuous segment length is used as the shadow acceptance basis quantity, and the shadow interaction recovery value is used to shrink the shadow acceptance basis quantity to form the shadow interaction confirmation quantity.

8. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 7, characterized in that, The step of selecting the target transmission path based on the shadow interaction confirmation includes, when the current competitive path set includes only one candidate cloud-edge transmission path, taking this candidate cloud-edge transmission path as the target transmission path. When the current competitive path set includes multiple candidate cloud-edge transmission paths, they are sorted according to the shadow interaction confirmation quantity. When the shadow interaction confirmation quantities of two candidate cloud-edge transmission paths are the same, the leading continuous segment length and the shadow interaction recovery value are compared in turn. When the shadow interaction confirmation quantity, leading continuous segment length and shadow interaction recovery value of two candidate cloud-edge transmission paths are all the same, the original sorting position of the two candidate cloud-edge transmission paths in the current competitive path set remains unchanged, forming a shadow interaction priority order, and the candidate cloud-edge transmission path ranked first in the shadow interaction priority order is taken as the target transmission path. The shadow interaction acknowledgment quantity, leading continuous segment length, and shadow interaction recovery value corresponding to the target transmission path are associated and recorded to form the target path record.

9. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 1, characterized in that, The establishment of a service transmission channel along the target transmission path and the parallel migration between the current transmission path and the target transmission path include searching for the record of the target transmission path in the candidate path record set and extracting the service session mapping status, access change relationship and cloud node change relationship. When the target transmission path is inconsistent with the current transmission path, a service transmission channel is established on the target transmission path according to the service session mapping status, access change relationship and cloud node change relationship. After the service transmission channel is established, the current service messages are synchronously sent to the current transmission path and the target transmission path to perform parallel migration, and the arrival time of the migration receipt returned to the service side device for each migration service message is recorded.

10. The low-latency interactive intelligent network selection method based on a hybrid cloud architecture as described in claim 9, characterized in that, The process of forming a new current transmission path includes comparing the arrival time of the migration receipt in the target transmission path with the arrival time of the real interactive receipt with the same message sequence number in the current transmission path one by one to form a migration receipt lead difference. The migration leading continuous segment length and migration recovery value are formed based on the migration receipt leading difference. The migration leading continuous segment length is used as the migration acceptance basis quantity. The migration recovery value is used to shrink the migration acceptance basis quantity to form the actual migration confirmation quantity. The actual migration confirmation is compared with the shadow interaction confirmation in the target path record. If the actual migration confirmation is not inferior to the shadow interaction confirmation, the target transmission path is replaced with the new existing transmission path.

Images