CN111432005A - A method for service migration under the condition of narrow-band weakly connected network - Google Patents

Abstract

The invention relates to a service migration method under the condition of a narrow-band weak networking, which comprises the following steps: (a) enabling each edge service node and the end user served by the edge service node to monitor the state of the logical connection with each other through periodic beacon information; (b) when the logic connection fails, the edge service node reports migration requirements to a virtual main node in a cloud center or a mobile edge service node set, and the end user actively searches other possible access objects and delegates the access objects to report to the cloud center or the virtual main node after successful access; (c) and the cloud center or the virtual main node determines a migration path and migrates the migration path according to the maintained and updated cloud-edge or edge-edge fusion network routing graph after acquiring the addresses of the initiating node of the service to be migrated and the receiving node expected to receive the migration service. Therefore, the migration efficiency of the service and the data is improved, and the application service is ensured to be uninterrupted.

Description

A method for service migration under the condition of narrow-band weakly connected network

technical field

The invention belongs to the technical field of network services, and relates to a method for service migration under the condition of a narrowband weak connection network, in particular to a method for improving the migration service and data efficiency under intermittent disconnection and narrowband weak connection network conditions.

Background technique

When the multi-node mobile information service center manages cross-node data, it needs to implement multi-node version synchronization updates for specific environments (intermittent weak connection characteristics and narrow-band weak connection networks) to solve the version consistency management problem caused by data distribution. Zhang Chunming proposed a timestamp-based replica consistency model RCMTS for massive data. The RCMTS model manages replicas through timestamp technology, and divides replica updates into out-of-domain updates and intra-domain updates through the high degree of autonomy of grid regions. This strategy improves the speed of update; and adopts the access strategy based on user view to ensure the correctness of user access to logical files, and proposes a dynamic and scalable copy positioning method DSRL, which uses index information nodes to support the same Simultaneous and efficient positioning of multiple copies of data, by using local index nodes to support local copy queries, a dynamic mapping technology is further proposed, which can allocate global copy positioning information according to the machine performance of index nodes, and supports index information nodes. dynamic joins and exits.

However, in order to ensure uninterrupted service when user equipment accesses across nodes, there are two uninterrupted service guarantee modes. One is the idea of container-based service lightweight and data unitization, user-centric service migration, the purpose of which is to migrate services and data to edge service nodes close to user equipment to obtain nearby services; the other is A relay channel is established between the new access point and the original access point to bridge the two, and the new access point acts as a relay between the user and the original access point. The former is beneficial to avoid business response timeout, but the migration time of services and accompanying data may be longer. Although the latter does not need to migrate services, the service response time will be restricted by the relay forwarding capability. Both of these migration-based assurance models face the question of how to improve the efficiency of migration services and data under intermittent disconnection and narrow-band weak network conditions.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method for service migration under the condition of narrow-band weak connection network in order to overcome the deficiencies of the prior art.

In order to achieve the above-mentioned purpose, the technical scheme adopted in the present invention is: a method for service migration under the condition of a narrow-band weakly connected network, which comprises the following steps:

(a) Make each edge service node and its service end users monitor the status of logical connections with each other through periodic beacon information;

(b) When the logical connection fails, the edge service node reports the migration requirement to the cloud center or the virtual master node in the mobile edge service node set, and the end user actively searches for other possible access objects and after successful access entrust it to report to the cloud center or the virtual master node;

(c) The cloud center or the virtual master node maintains and updates the cloud-edge or edge-edge fusion network route according to the obtained addresses of the initiating node of the service to be migrated and the receiving node of the service expected to receive the migration service. Figure, determine the migration path and carry out the migration.

Optimally, in step (a), the access status information of the end user is periodically reported to the virtual master node in the cloud center or the mobile edge service node set through the edge service node to which it is connected.

Further, in step (a), the access state information includes the identification number of the end user and the current location of the end user, the identification number of the accessed edge service node, the current location of the edge service node, and the number of successful accesses. The information content of the access times includes the logical connection maintenance time of each access and is listed in the chronological order of the access requests.

Further, in step (a), the cloud center or the virtual master node calculates the metric values of the access request density and the average access maintenance time for each end user, and uses the historical access sequence diagram to represent; the request density It is the number of times that the access object is discovered and the request is sent within a given time period to measure the frequency of an end user encountering an edge service node. The average access maintenance time is the connection maintenance of all successful accesses within a given time period Average duration to measure the stability of the connection.

Further, in step (a), the cloud center or virtual master node is based on the k-means clustering algorithm. The distances of the class centers are clustered as metric values, resulting in k clusters.

Optimally, in step (c), the migration adopts the service lightweight containerization package migration technology.

Due to the application of the above technical solutions, the present invention has the following advantages compared with the prior art: the method for service migration under the condition of a narrow-band weak connection network of the present invention, by monitoring the logical connection between the edge service node and the end user, when the edge service node has migrated When required, the cloud center or virtual master node determines the addresses of the initiating node of the service to be migrated and the receiving node that expects to receive the migration service, and migrates according to the converged network routing map, thereby improving the efficiency of service and data migration and ensuring that application services are not Intermittent.

Description of drawings

1 is a schematic structural diagram of a container image layered framework of the present invention;

Fig. 2 is the user, service and data interdependence relationship diagram of the present invention;

Fig. 3 is the data rights management diagram of the present invention based on user, service, data interdependence;

4 is a migration diagram of a data unit and a lightweight service according to the present invention;

FIG. 5 is a framework diagram of an intelligent decision-making for cross-node access and migration in a wireless communication environment of the present invention.

Detailed ways

The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

The method for service migration under the narrowband weak connection network condition of the present invention comprises the following steps:

(a) Make each edge service node and its service end users monitor the status of logical connections with each other through periodic beacon information.

The access status information of each end user is periodically reported to the virtual master node in the cloud center (referred to as cloud) or mobile edge service node (referred to as edge) set through the edge service node it accesses. This type of status information includes the end user identification number (or address) and current location, the access edge service node identification number (or address) and current location, and the number of successful accesses (including the logical connection maintenance time of each access, according to the chronological order in which access was requested). Based on this kind of state information, the cloud center or virtual master node calculates the following metrics for each end user: access request density (the number of times that access objects are found and requests are issued within a given period of time, which is used to measure the chance that an end user encounters a certain The frequency of edge service nodes), the average access maintenance time (the average value of the maintenance duration of all successfully accessed connections within a given time period, used to measure the stability of the connection), and represented by the historical access sequence diagram, that is, the abscissa value It is the reporting time of status information, and the ordinate value is these two status values.

The virtual master node in the cloud center or mobile edge service node set is based on the k-means clustering algorithm, with the number of edge service nodes as the parameter k, the coordinates of the edge service nodes as the initial cluster center, and the distance from the end user to the cluster center. Clustering for metric values yields k clusters. According to the distance between the edge service node in the cluster and the center point, it is divided into three levels: "near", "medium" and "far", and the number of end devices in the cluster is divided into "small", "medium" and "large" At the third level, these two dimensions are used to comprehensively measure the potential access load of edge service nodes. Nine potential access load levels can be formed by combining them. For example, if the number of end users in the cluster where the edge service node that is “close” to the clustered center is located is “large”, then the potential access load of the edge service node is “large”. The input load may be the heaviest. The potential access load of each edge service node is also represented as a potential load timing diagram.

Based on the above two types of sequence diagrams, decisions for service migration can be made. For example, after determining the source edge service node and the destination edge service node to be migrated, whether to actually implement the migration needs to make the following judgment: First, check the potential load sequence diagram of the destination edge service node. If it is judged that the load is heavy, it is not appropriate to accept the migration. If it is judged that it can be received, check the historical access sequence diagram of the destination edge service node for the service object of the migration service (that is, an end user). If it is judged that the requested access density is extremely low or the average access maintenance time is extremely short, it is not appropriate Migration; if it is judged that the migration is suitable, check the end user's historical access sequence diagram of the source edge service node. If it is judged that the requested access density is extremely low or the average access maintenance time is extremely short, the retained service copy will not be considered. In order to improve the accuracy and timeliness of time sequence diagram analysis, principal component analysis, independent component analysis, principal feature analysis and logistic regression can be used for combined analysis.

(b) When the logical connection fails, the edge service node reports the migration requirement to the cloud center or the virtual master node in the mobile edge service node set, and the end user actively searches for other possible access objects and entrusts them to the report from the cloud center or the virtual master node;

(c) The cloud center or the virtual master node maintains and updates the cloud-edge or edge-edge fusion network route according to the obtained addresses of the initiating node of the service to be migrated and the receiving node of the service expected to receive the migration service. Figure, determine the migration path and carry out the migration.

If the cloud center or the virtual master node that replaces its work can gain insight into the access behavior trajectory of end devices (ie end users), it is helpful to formulate a reasonable post-migration service copy retention strategy and a general basic service cache deployment strategy. The differentiated parts of the service to be migrated are determined according to the operating environment of the service at the migration destination, and are reasonably divided into basic units suitable for transmission, which are piggybacked into the IP packet flow of the same transmission path to reduce network bandwidth occupation. If you decide to keep a copy of the service, you should perform a "downsizing" process to reduce resource usage, including splitting the service, recording the service component list and assembly process, retaining differentiated components, and deleting common components that can be copied locally.

The purpose of lightweight service is to reduce the amount of transmission during migration to adapt to intermittent disconnection and narrowband weak connection network environment. At the same time, it should have a fast startup feature to quickly resume application services suspended due to migration. Learn from container technology, study the image layering mechanism and the basic environment image preloading strategy, and adopt a lightweight containerized packaging and migration solution that adapts to the highly dynamic set of edge service nodes, intermittent disconnection and narrow-band weak network conditions (ie, service lightweight The high-level containerized packaging and migration technology, as shown in Figure 4), reduces the network bandwidth consumption of the service image, improves the migration speed and quickly restores the service.

A deliverable software product developed based on container technology can be regarded as a container application service image, that is, an application built based on a hierarchical relationship. From bottom to top, it includes a boot file system (such as bootfs), a root file system (such as rootfs) ), development environment tools (such as emacs, apache), and applications. Each image layer has a corresponding json file, the purpose of which is to provide information such as what process should run on top of the image, and what environment variables should be configured for the process. To deploy the container application service image and provide services, the container daemon needs to create a corresponding container application service container according to the image json file, that is, generate a dynamic container according to the static image. When the container starts, a container layer will be added on top of the container application service image to run one or more processes. Each process occupies the corresponding memory, CPU, virtual network device resources, and the image layer file of the container image. Provided file system resources. The process in the container writes a task file according to the Copy-on-Write mechanism, that is, the file is first copied from the container image layer to the top container layer, and then the relevant process writes to the copy in the container layer. . It can be seen that the new or modified data is stored in the top container layer, while the data of the mirror layer remains unchanged. Based on the reasonable layering of images, building a new application image is simplified to reuse public images and add application-related business logic programs. Combined with the incremental upload and download mechanism of the mirror warehouse, new applications can be quickly uploaded to the mirror warehouse, and then quickly distributed to other nodes.

This application proposes a container layering mechanism, and the container public image is divided into a system kernel layer, an operating system layer, a common component layer, a development language layer, and a development framework layer from bottom to top. Developers can choose any number of image layers from bottom to top on the left in Figure 1 as their base image, and develop application service images based on this. In this way, you can take full advantage of the container layering feature to maximize the utilization of public images. For a specific application service container, only the content of the application logic layer and service container layer at the top of the image is unique to the service, and other lower images can be considered as the public image layer supporting the service. Therefore, when you need to migrate an application service, you should know in advance the adaptation of the destination's public base image. If it is fully adapted, only the content of the application logic layer and service container layer specific to the service needs to be migrated to reduce the transmission amount during migration (the content of the service container layer may include process recovery information, environment variable configuration requirements, application business Intermediate results, accompanying data, etc., can be a json-like file in form). If there is an incompatibility, the incremental synchronization operation can be performed for the public basic image, or the pre-migration of the public basic image can be guided based on the big data analysis results of the end user's access behavior at will. Also inform the destination to preload the public base image. If the common base image requires incremental synchronization, it should take precedence over the content migration of the application logic layer and the service container layer to ensure the timely completion of preloading (that is, the content at the application logic layer and service container layer reaches the destination done before), thereby speeding up service recovery.

In order to meet the requirement of "services and data on-the-go", it is necessary to effectively manage the multiple types of data to be migrated, quickly extract necessary data, and lay the foundation for migrating and restoring services as soon as possible. The multiple types of data to be migrated include service container layer process recovery information, environment variable configuration requirements, intermediate results of application services, and data of various types and formats that are carried around, which increases the difficulty of data management. In order to meet the requirements of various service and data migration scenarios and information synchronization based on data units, research on data unit management and on-demand rapid extraction technology is expected to reduce the difficulty of data management.

The essence of a data unit is a dynamic association and mounting service between users, services, and data. According to the type of data or data set contained in the data unit, the data unit can be divided into different types, and the degree of dependence on the data (such as strong dependence, weak dependence, etc.) is recorded in the data unit. The composition of the data unit can be a single data, a single file, a database table, an object, and so on. Data units are also aggregatable and nestable at the same time. For example, a directory composed of multiple files, or a data unit collection of multiple database tables plus multiple files can form a new data unit. The same file is allowed to be a constituent element of multiple data units. The content of a data unit is allowed to dynamically change vertically, but not horizontally. If horizontal changes are required, only one more data unit based on this data unit can be created. Storage services such as file systems, relational databases, and message queues can also be managed by data units. Users can obtain various application services by accessing the edge information service system. In the event of random access behavior, services need to be migrated in order to provide services to closer access points. This involves the migration of related multi-type, multi-format, and multi-source data, which can be included in a data unit, which facilitates unified packaging, unified migration, and rapid recovery of services and data.

The creation time of data units is not limited and can be initiated on demand at any point in the data life cycle. For example, in the data collection period, the collector can divide the data units according to the data source, data space-time attributes, data collection methods, etc.; in the data storage period, the storage initiator can according to the destination of the data landing, such as a certain directory, a certain The data unit is divided into a database and a certain table, thus increasing the flexibility of data storage management; in the data processing period, the processor can declare the data set as a data unit during the data processing process, and then use various operators in On the data unit, the speed of the data processing process is accelerated; in the data distribution and sharing period, the owner performs the data distribution, data sharing and data subscription process in units of data units. Therefore, when a service needs to be migrated, first, according to the content characteristics of the application logic layer and the service container layer, a suitable content storage form (for example, a storage form that can adapt to the rapid recovery service) should be adopted, and the state of the intermediate result of the current business processing should be adopted. The mode determines its storage form; then quickly extracts necessary accompanying data or data sets according to the application business logic; finally, based on the determined storage forms of various types of data, data units are created to carry the above-mentioned various types of migration data.

The interdependence between users, services and data is shown in Figure 2. Users with legal identities can only obtain access rights to "non-data access services" based on authenticated identities, such as sending their own information, submitting perceived environmental information, asking for non-confidential information, etc. However, when it needs to access the "data access service", it must rely on the authenticated identity and the obtained authorization token. "Data access services" involve operations on data, and therefore, must pass identity and permission verification to be allowed. First, the identity of a user who uses a "data access service" is legal and has the right to use the "data access service". Then, the identity of the "data access service" is also legal, and there is a right to use the "data access service". The right to operate or process a certain type of data (for example, user collection data, service processing data, system storage data, etc.).

As shown in Figure 3, after requesting registration, submitting audit information, and getting feedback that the audit is passed, the user has a legal identity in the edge information service system. Based on the legal identity of the user and the application service request submitted by it, after analyzing its application business logic, the system will grant it appropriate permissions to ensure that it can obtain satisfactory services without causing other users or system permissions. violated. Like users, any application deployed to the system must be certified in advance to ensure traceability of the source. At the same time, it needs to be granted appropriate data permissions to ensure that the data is used safely. According to the source of the data entering the system, the collector or publisher gives the permission requirements for the data to be accessed by the application service. The data permission management service of the system will perform unified permission setting, update, and release operations on the data. For the data collected by the user, the user will process the data in a unitized manner according to the collection time, location, collection method, etc., and attach the data access policy accordingly for reference when the data rights management service is processing. During the operation of the application service, if the intermediate results generated need to be shared with relevant partners, a data unit can also be created based on the shared content, and a data access policy can be attached accordingly for reference by the data rights management service. The data stored in the system itself is also managed by the data rights management service. According to the application service's requirements for data types, data formats, data processing methods, etc., data unit processing is performed, and reasonable access rights are set to meet the application service's requirements for data. On the premise of access requirements, ensure the security of the data itself.

Under the network conditions of ensuring network connectivity and uncertain wireless network conditions, research services and data cross-node access and migration technologies, using "computation-intensive", "data-intensive", "delay-sensitive", " The characteristics of five dimensions, such as fault-tolerant control type and frequent interaction type, characterize the characteristics of application services in a mobile environment, and then describe the technical approaches of cross-node access and migration of services and data in typical combination scenarios.

The main characteristics of "computing-intensive" applications: need to consume a lot of CPU resources; usually a computing task occupies a large number of computing nodes at the same time; the number of CPU cores should match the number of concurrent subtasks; typical applications include high-definition video decoding, deep learning training process, etc.

The main characteristics of "data-intensive" applications: a large number of independent data analysis and processing tasks are distributed and processed on different loosely coupled computing nodes; it has highly intensive mass data I/O throughput requirements; usually has data flow-driven processes; typical applications There are Web applications, software-as-a-service cloud facilities, and information systems that have high demands on the ability to acquire, manage, analyze, and understand massive and rapidly changing data. Such applications need to rely on data-intensive computing capabilities, including high-performance computing capabilities, data analysis and mining capabilities.

Main characteristics of "delay-sensitive" applications: Most applications have certain requirements for delay, and usually have a threshold that can be tolerated. When this threshold is small, high-priority resource allocation is required to ensure it. Due to the limited resources, the difficulty of guarantee increases, the probability of not being able to be effectively guaranteed increases, and the user's application experience is very poor. "Latency-sensitive" applications refer to applications that require a very small delay threshold to meet business requirements. Services that are "latency-sensitive" are also sensitive to temporary interruptions in logical connections.

The main features of "fault-tolerant control" applications: The fault-tolerant control here refers to that when some functional components in the information system fail, the information system can still provide services according to the expected performance or under the condition of acceptable performance loss. Services with a "fault-tolerant control" feature will be more tolerant of temporary interruptions in logical connections.

The main features of "frequently interactive" applications: usually involve the execution of IO-intensive tasks such as network transmission and disk IO, but consume very little CPU resources, and most of the time of the task is waiting for the IO operation to complete. Typical applications include Web applications, etc.

The application business in the mobile environment may include obtaining comprehensive intelligence data and receiving task instructions from the cloud center, pushing complex computing tasks to the cloud center for execution, deploying services supported by massive data in the cloud center and providing nearby access access on the edge platform, The edge platform integrates multiple end-to-end data for sharing and comprehensive decision-making, the cloud center performs computationally intensive intelligent training tasks, and the edge makes inference based on the training model to meet the needs of "delay-sensitive" applications, following the order of end, edge, and cloud. Data is graded from coarse to fine to ease transmission pressure, and services are cached step by step in the order of cloud, edge, and end to reduce access latency.

In view of the characteristics of business applications in a mobile environment, the actual service types are rich and diverse, and the network conditions are highly dynamic. Therefore, cross-node access and migration technologies for services and data that adapt to different service characteristics under different network conditions are needed, and an integrated cross-node access and migration mechanism that is transparent to relevant access requesters is needed. The key breakthroughs are the transparent cross-node access and migration framework for adaptive service characteristics under cloud-edge and edge-edge communication conditions, and the transparent cross-node access and migration framework for adaptive service characteristics under uncertain network conditions.

The intelligent decision-making framework for cross-node access and migration in the wireless communication environment is shown in Figure 5 (the edge node with the highest probability of opportunistic connection with other connected subnets is selected to replace the cloud center to perform coordination duties, mainly for When two connected subnets are connected, it is convenient for them to use a higher probability of connection than other edge nodes to perform access authentication and meta-information synchronization. In this intermittent weak connection environment, it is necessary to ensure that services and data are separated between the isolated subnets. Pre-synchronization is performed between them. Therefore, it is very important to make functional components and data units, as well as to accurately pre-set basic common functional components and common data units. This helps to build a lightweight migration service container image, and at the same time, it is also convenient for The migration service only carries personalized incremental data units to avoid unnecessary redundant migrations). Under this framework, the edge processing engine set on the edge computing platform is responsible for receiving service requests from end devices. Whether the edge computing platform independently decides to establish a service for the terminal device or requests the cloud center to collaborate and build a service, it will redirect the service request to the cloud processing engine of the cloud center, and perform intelligent analysis with the support of massive data and powerful computing power of the cloud center and decisions in order to get the connection maintenance strategy. For example, if an end device leaves the currently accessed edge node and accesses other edge nodes as it encounters, whether to use edge-edge collaboration-based service migration or data migration, or cloud-edge collaboration service and data migration, it needs to be intelligently decision to decide. The intelligent analysis and decision-making module works based on the trained machine learning model, which can also be pushed forward to the edge computing platform to reduce the response delay. The end device will also entrust the currently connected edge node to feed back the random access experience to the cloud center. The cloud center executes machine learning to train the decision-making model according to various feedbacks received, combined with service characteristics and dynamic changes in network conditions.

The machine learning process is modeled using the Q-Learning method. The method includes a reward table and a Q table, both of which can be represented as a two-dimensional matrix with states as rows and actions as columns, but the value of the reward table is the payoff of taking an action in a given state, and the value of the Q table Action preferences are recorded. For services with characteristics of "computation-intensive", "data-intensive", "delay-sensitive", "fault-tolerant control" and "interaction-frequent", analyze their performance in maintaining effective connection based on edge-to-edge service migration. The performance of the logical connection in the entry environment, the performance of the logical connection in the effective access environment based on edge-edge data migration, and the performance of the logical connection in the effective access environment based on cloud-edge collaboration. The analysis result is used as the initial value of the reward table in Q-Learning. This performance value is the application performance conformity of the end-to-end transmission constrained by the delay on the logical connection. From the perspective of the end device experience, non-conformity, basic conformity, and conformity are used to qualitatively measure.

The state set in Q-Learning can consider the dimensions that reflect the characteristics of the service, such as "computation-intensive", "data-intensive", "delay-sensitive", "fault-tolerant control" and "interaction frequently", as well as the end requesting service. The frequency of device cross-node access (such as frequent cross-node access, occasional cross-node access, and no cross-node access) is obtained by combining them. The types of actions in Q-Learning include the strategy of using edge-to-edge service migration to ensure access availability, the strategy of using edge-edge connection relays to ensure access availability, the strategy of using cloud-edge collaboration to ensure access availability, and Do not adopt any strategy to guarantee the validity of access.

When the cloud center receives the service request redirected by the edge processing engine, it obtains the characteristics of the requested service through the intelligent analysis and decision-making module, characterizes it with one of the five types of dimensions or a combination thereof, and then according to the history of the end device access behavior The data is mined and analyzed by typical data mining methods, and the frequency category of cross-node access is obtained, thereby obtaining the current state of the terminal equipment. Based on the trained Q table, a feasible path is searched, and the current state is converted to a state that can satisfy the logical connection delay constraint end-to-end transmission application performance consistency, and the action that triggers the satisfactory state is the desired connection maintenance strategy. The timely feedback of the actual experience of the end device and the design of the Q value update function are also related to the stability of the end-to-end transmission application performance constrained by the delay of the logical connection.

In a mobile environment, the probability of the network being fully interconnected at the same time gradually decreases as the environment deteriorates. At any time, there is one or several isolated edge-edge interconnection subnets, and the possibility of even completely isolated edge nodes gradually increases. At the same time, isolated edge-edge interconnection subnets or completely isolated edge nodes also exhibit high dynamics, with time-to-time aggregation and interconnection and separation and disconnection. On the one hand, using different movement models (such as random waypoint, random walk, random direction, random travel, smooth random movement, highway movement, Manhattan grid movement), the dynamic change trend of network topology under different node scale scenarios is discussed. Cloud-edge subnets and edge-edge subnets maintain network conditions for connectivity, model opportunistic connectivity probability models, and guide the design of intelligent decision-making frameworks for cross-node access and migration in uncertain wireless environments. On the other hand, the "virtual heartbeat" mechanism is adopted, which requires each pair of nodes in the cloud-edge fusion network to monitor the connectivity information with each other. In order to save network bandwidth and make full use of application service data packets for piggybacking, the aforementioned streaming migration technology based on opportunistic piggybacking and meta-information synchronization load minimization technology oriented to service characteristics and opportunistic piggybacking can also be used to transmit heartbeat packets. Based on the information exchanged by heartbeat packets, edge nodes isolated from the cloud can be connected to each other to form an independent connected subnet, and the edge node with the highest probability of opportunistic connectivity with other connected subnets is selected to replace the role of the cloud center to coordinate this connected subnet. The cross-node access connection of the internal end device works on occasion, that is, in this connected subnet, it provides a route between the current access point and the access point disconnected due to movement for the end device connected across the nodes. Assists in access authentication and meta-information synchronization.

The above-mentioned embodiments are only intended to illustrate the technical concept and characteristics of the present invention, and the purpose thereof is to enable those who are familiar with the art to understand the content of the present invention and implement them accordingly, and cannot limit the protection scope of the present invention. All equivalent changes or modifications made according to the spirit of the present invention should be included within the protection scope of the present invention.

Claims (6)

1. A method for service migration under narrowband weak networking conditions, comprising the steps of:

(a) enabling each edge service node and the end user served by the edge service node to monitor the state of the logical connection with each other through periodic beacon information;

(b) when the logic connection fails, the edge service node reports migration requirements to a virtual main node in a cloud center or a mobile edge service node set, and the end user actively searches other possible access objects and delegates the access objects to report to the cloud center or the virtual main node after successful access;

(c) and the cloud center or the virtual main node determines a migration path and migrates the migration path according to the maintained and updated cloud-edge or edge-edge fusion network routing graph after acquiring the addresses of the initiating node of the service to be migrated and the receiving node expected to receive the migration service.

2. The method of service migration under narrowband weak networking conditions of claim 1, wherein: in the step (a), the access state information of the end user is periodically reported to a virtual main node in a cloud center or a mobile edge service node set through the edge service node accessed by the end user.

3. The method of service migration under narrowband weak networking conditions of claim 2, wherein: in the step (a), the access state information includes an end user identification number and an end user current position, an accessed edge service node identification number and an edge service node current position, and successful access times, and the information content of the successful access times includes logic connection maintaining time of each access and is listed according to the time sequence of the requested access.

4. A method of service migration in narrowband weak networking conditions according to claim 2 or 3, characterized by: in the step (a), the cloud center or the virtual main node calculates the measurement values of the access request density and the average access maintenance time for each end user, and the measurement values are represented by a historical access time sequence diagram; the request density is the frequency of finding access objects and sending requests in a given time length to measure the frequency of one end user to encounter a certain edge service node, and the average access maintenance time is the average value of all successfully accessed connection maintenance time lengths in the given time length to measure the stability of connection.

5. The method of service migration under narrowband weak networking conditions of claim 2, wherein: in the step (a), the cloud center or the virtual master node performs clustering based on a k-means clustering algorithm by using the number of edge service nodes as a parameter k, edge service node coordinates as an initial clustering center, and a distance between an end user and the clustering center as a metric value, so as to generate k clusters.

6. The method for service migration in narrowband weak networking conditions of claim 1, wherein: in the step (c), the migration adopts a service lightweight containerization package migration technology.

Images