WO2022120959A1

WO2022120959A1 - Load balancing multi-service migration method and system based on minimum-cost maximum-flow

Info

Publication number: WO2022120959A1
Application number: PCT/CN2020/138815
Authority: WO
Inventors: 唐欢; 王洋; 张锦霞; 须成忠; 叶可江
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2020-12-10
Filing date: 2020-12-24
Publication date: 2022-06-16
Also published as: CN112601232B; CN112601232A

Abstract

The present invention relates to the technical field of information. Disclosed are a load balancing multi-service migration method and system based on a minimum-cost maximum-flow. Said method comprises: constructing a network topology map according to the geographic position and connection relationship of edge servers; calculating the shortest path distance between each pair of edge servers, and generating a shortest path distance matrix; establishing a minimum-cost maximum-flow model, solving the model to obtain a minimum-cost maximum-flow, and placing a service node on a corresponding edge server node; updating access information of each service, adjusting the weight of the edge, connected to a terminal node, of the edge server node, and updating the position of the service node on the edge server node; and monitoring and collecting statistics of the access information in real time, and when the rate of change of the access information exceeds a set threshold, adjusting the service node and edge server node corresponding thereto. The present invention allows for more balanced, flexible and adjustable placement of a virtual service on an edge node, thereby guaranteeing quick response to a user and improving the service quality.

Description

Multi-service migration method and system based on load balancing with minimum cost and maximum flow

technical field

The present invention relates to the field of information technology, and more particularly, to a multi-service migration method and system based on load balancing with minimum cost and maximum flow.

Background technique

Mobile edge computing (MEC) is a new computing paradigm that combines the advantages of mobile computing and edge computing to improve the quality of service (QoS) for mobile users. With MEC in particular, computing resources can be pushed from the cloud center to the edge of the network, which enables data services and other related processing tasks to run close to mobile users. Therefore, it not only reduces service latency, but also minimizes network traffic, both benefits of which are quite important for those time-bound services (typical applications in mobile computing).

Given its advantages of access efficiency and low cost, MEC is quickly becoming the focus of the next wave of research, especially in 5G networks with smarter and more powerful computing resources, which can be installed at the edge of wireless networks, enabling many time-sensitive services deployment is closer to the user. However, considering the mobility of users, the time difference of access to services, and the limited coverage of a single edge server, if multiple services are fixed on one edge server at the same time, there will be resource contention and delay in accessing services. Increase. Therefore, if these factors are not taken into account, the service provided may significantly increase the access delay, and worse, cause a large amount of network traffic, cause network congestion, lead to service performance degradation, or even service interruption.

With the rapid development of virtualization technology in cloud computing, more and more application services are encapsulated in virtual machines or containers to provide users with on-demand mobile services in MEC, which can effectively alleviate the problems proposed in the previous paragraph. question. In many scenarios, a large number of mobile users access services at different times and places. Therefore, service placement and migration strategies cannot generally be designed for each user, but depend on the time-space dimension of mobile access patterns. Dynamic changes. At the same time, the migration of services is not free, the cost is the cost of large data transfer, and may cause service interruption, thereby increasing the overall service delay. Therefore, the goal of migration is usually to reduce migration cost and service latency, while having good performance guarantees.

Traditionally, dynamic programming algorithms, heuristics, and machine learning methods are often used to solve this problem. But these methods all have high time complexity, space complexity and offline properties. Therefore, when the network scale is large, the solution time of the strategy is too long, which cannot meet the low-latency requirements of the application. These methods are also difficult to cope with sudden increases in requests and abnormal situations when users visit intensive times and places. At the same time, these methods can also ignore the communication cost of service migration.

In contrast, some online transfer algorithms can trade off between accuracy and response time. Some scholars have used reinforcement learning to improve the existing problems to a great extent, but it is only aimed at a single service, and most scenarios of MEC involve multiple services. Some people use the Markov model, but it is aimed at the dynamic service problem in mobile cloud services, which can derive the best migration strategy, but does not consider the problem of balanced placement. Some scholars have formulated the same problem as a one-to-one contract game model, and developed a learning-based price control mechanism to effectively deal with the resources of MEC. By applying the game method and learning process, the method can acquire the dynamic information of the MEC system through continuous interaction with the unknown system environment. Unfortunately, it always takes a while to learn, and before it converges, the quality of decision-making cannot be guaranteed, and low latency of service cannot be achieved.

Existing work has achieved good results in the migration strategy, but it often does not consider the problem of resource contention for multiple services due to limited edge server resources, and with the increasing demand of people, the network scale also increases. The access patterns are also more complex, and the previous strategies are difficult to make optimal decisions under large-scale network topologies. At the same time, the previous strategy was either for the migration of a single service, or lacked a certain degree of dynamism, or the solution time was too long. The main disadvantage is that it cannot take into account low service delay and load balancing of each edge node at the same time.

technical problem

The purpose of the present invention is to provide a multi-service migration method and system based on load balancing with minimum cost and maximum flow in view of the technical problems existing in the prior art, so that the placement of virtual services on edge nodes is more balanced and flexible and adjustable .

technical solutions

In order to solve the problem proposed above, the technical scheme adopted in the present invention is:

The present invention provides a multi-service migration method based on load balancing with minimum cost and maximum flow. The specific steps of the method include the following steps:

Build a network topology map based on the geographic location and connection relationship of edge servers;

Calculate the shortest path distance between each pair of edge servers and generate a shortest path distance matrix;

According to the distance matrix, combined with the access information of each service, the carrying capacity of each edge server and its relationship with the service delay, a minimum cost maximum flow model is established, and the model includes source nodes, service nodes, edge server nodes and endpoints;

Solve the model to obtain the minimum cost and maximum flow, and place the service node on the corresponding edge server node;

Update the access information of each service, adjust the weight of the edge connecting the edge server node to the endpoint, and update the position of the service node on the edge server node to achieve load balancing;

Real-time monitoring and statistics are performed on the access information of each service. When the change rate of the access information exceeds the set threshold, the corresponding service nodes and edge server nodes are adjusted.

Further, the minimum cost maximum flow model is specifically:

Taking the S node and the T node as the source node and the end node of the model, Ni represents the ith service node, Ej represents the jth edge server node, the source node S is only connected to the service node Ni, each Each service node Ni is connected to all edge server nodes Ej, and each edge server node Ej is connected to the termination node T; each edge in the model is associated with capacity and communication cost.

Further, each edge server node Ej is connected to the terminal T through at least two edges according to the number of services it can bear, and the cost weights of each edge are different and in an increasing relationship.

Further, the specific process of updating the position of the service node on the edge server node includes:

In the residual network of the model solving algorithm, update the access cost and migration cost of each service node placed on the edge server node;

According to the updated access information, delete and add corresponding service nodes and edge server nodes in the residual network;

Adjust the weight of the edge connecting the edge server node to the endpoint according to the current edge network index changes and the load of each edge server node;

The incremental minimum cost and maximum flow algorithm is used to solve the problem, and the position of the service node on the edge server node is updated according to the obtained minimum cost and maximum flow.

Further, the incremental minimum cost maximum flow algorithm specifically includes the following:

The access volume of the service node changes, and when the variable of access times is greater than the pre-learned threshold, the location of the service node is updated;

When the number of service nodes changes, delete service nodes or add service nodes, and update the location of service nodes;

When the capacity of the edge server node changes, and when the capacity of the edge server node decreases and the flow from it is larger than the reduced capacity of the edge, or when the edge server node is deleted, and when the capacity of the edge server node increases Or when a new edge server node is added, the location of the service node is updated.

Further, starting from the changed node, re-find a new minimum cost maximum flow, migrate and place the service node, and update the position of the service node on the edge server node.

The present invention also provides a load balancing multi-service migration system based on minimum cost and maximum flow, the system includes the following:

Topology map building module: used to build a network topology map according to the geographical location and connection relationship of edge servers;

Matrix generation module: used to calculate the shortest path distance between each pair of edge servers in the network topology diagram, and generate a shortest path distance matrix;

Model building module: used to establish a minimum-cost maximum flow model according to the distance matrix, combined with the access information of each service, the load capacity of each edge server and its relationship with the service delay, and the model includes the source node , service nodes, edge server nodes and edge nodes;

Service node migration module: used to solve the model to obtain the minimum cost and maximum flow, migrate the service nodes, and place them on the corresponding edge server nodes;

Service node update module: used to update the access information for each service, adjust the weight of the edge connecting the edge server node to the endpoint, and update the position of the service node on the edge server node;

Monitoring and adjustment module: It is used to monitor and count the access information of each service in real time, and adjust the corresponding service nodes and edge server nodes when the change rate of the access information exceeds the set threshold.

Further, the service node update module includes:

Cost update sub-module: used to update the access cost and migration cost of each service node placed on the edge server node in the residual network of the model solving algorithm;

Node deletion and addition sub-module: It is used to delete and add service nodes and edge server nodes in the residual network according to the updated access information;

Weight adjustment sub-module: adjust the weight of the edge connecting the edge server node to the endpoint according to the current edge network index changes and the load of each edge server node;

Node update sub-module: use the incremental minimum cost and maximum flow algorithm to solve, and update the position of the service node on the edge server node according to the minimum cost and maximum flow obtained by the solution.

The present invention also provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of any one of the methods when the processor executes the program.

The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the steps of any one of the methods are implemented.

beneficial effect

Compared with the prior art, the beneficial effects of the present invention are:

The multi-service migration method and system provided by the present invention transform the communication cost problem and load balancing problem of accessing services in the edge network into the problem of minimum cost and maximum flow, and can realize that only a small part of the access cost is increased. The placement on the edge nodes is more balanced and flexible, and it can also well cope with the sudden increase in the number of service accesses and abnormal situations. It is simple, reliable and easy to implement, ensuring fast response to users. Improved service quality.

Description of drawings

In order to illustrate the solutions in the present invention more clearly, the following will briefly introduce the accompanying drawings used in the description of the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present invention, which are common in the art. As far as technical personnel are concerned, other drawings can also be obtained based on these drawings without any creative effort. in:

FIG. 1 is a flowchart of a multi-service migration method for load balancing according to the present invention.

FIG. 2 is a network topology diagram of virtual service migration according to the present invention.

FIG. 3 is a schematic diagram of the minimum cost maximum flow model of the present invention.

FIG. 4 is a flow chart of the present invention for updating the location of a service node.

FIG. 5 is a schematic diagram of the last residual network of the present invention.

FIG. 6 is a schematic diagram of the modified residual network of the present invention.

FIG. 7 is a schematic diagram of a multi-service migration system for load balancing according to the present invention.

FIG. 8 is a schematic diagram of a service node update module in the multi-service migration system of the present invention.

FIG. 9 is a schematic diagram of the principle of the electronic device of the present invention.

BEST MODE FOR CARRYING OUT THE INVENTION

In order to facilitate understanding of the present invention, the present invention will be described more fully hereinafter with reference to the related drawings. Preferred embodiments of the invention are shown in the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that a thorough and complete understanding of the present disclosure is provided.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terms used herein in the description of the present invention are for the purpose of describing specific embodiments only, and are not intended to limit the present invention.

Referring to FIG. 1, the present invention provides a load balancing multi-service migration method based on minimum cost and maximum flow, and the specific steps of the method include the following:

Step S1: Build a network topology map according to the geographic location and connection relationship of the edge server. Specifically, the weight of the edge in the network topology graph is determined by the distance between the connected edge nodes and the network transmission speed.

The application scenario of the present invention can be abstracted into a network topology diagram as shown in Figure 2. In the edge network, the virtual service runs on the server of the edge node, so the edge node is abstracted into the edge server node, that is, Figure 2 Among the edge servers 1 to 5 connected to the decision center, users can access the corresponding edge servers for the services they need through the nearest access points 1 to 5.

Step S2: Calculate the shortest path distance between each pair of edge servers in the network topology diagram, and generate a shortest path distance matrix. Specifically, using the shortest path algorithm to calculate the shortest path distance, the method is simple and reliable.

Step S3: Count the access information of each service in the most recent time period, and establish a minimum cost maximum flow model according to the shortest path distance matrix and in combination with the load capacity of each edge server and its relationship with the service delay. The model includes source nodes, service nodes, edge server nodes, and endpoints.

Specifically, according to the shortest path distance matrix, the access cost and migration cost of each service placed on each edge node are calculated, and the relationship between the load capacity of each edge server and the service delay is calculated, and a minimum cost maximum flow model is established. .

In the embodiment of the present invention, referring to FIG. 3 , the minimum cost maximum flow model is specifically:

Take the S node and the T node as the source node and end node of the model, which are set for building the model. Ni represents the ith service node, and Ej represents the jth edge server node, where i is 1 to n, and j is 1 to m, that is, there are a total of n service nodes and m edge server nodes. The source node S is only connected to the service node Ni, each service node Ni is connected to all edge server nodes Ej, and each edge server node Ej is connected to the termination node T. Each edge in the model is associated with two values, capacity and communication cost.

Specifically, the capacity of each edge connecting the source node S and the service node Ni is 1, so that all services can be placed at the edge server or cloud center, and the cost is 0. The capacity of the edge connecting each service node Ni with all edge server nodes Ej is 1, and the cost is the sum of the delay cost and the migration cost of placing the service node on the edge server node Ej.

Further, each edge server node Ej is connected to the terminal T through several edges (that is, at least two edges) according to the number of services that the edge server node can carry, and each edge has a different cost weight. and an increasing relationship. Specifically, the weight of each edge can be determined according to the delay and cost of placing a corresponding number of services in the past, or it can be determined according to the user's needs or the real-time load of the entire edge network. If the weights of each edge are all set to 0, the result is the theoretical minimum total access delay, but it often occurs that the number of services that need to be placed on some edge nodes is particularly large, while other edge nodes are The number of services that need to be placed is relatively small, or even zero, which is likely to cause the edge nodes with heavy load to face resource contention, which will increase the delay of the service, or even cause downtime, which will degrade the quality of service.

In the embodiment of the present invention, the services on the edge server nodes with heavy load are placed on the relatively idle edge server nodes, and the total service access delay will not increase too much. By adding multiple edges on the edge connecting the edge server node to the endpoint, that is, splitting one edge from each edge server node to the endpoint into multiple edges, and the number of edges is the number that can be placed by the corresponding edge server node. The number of services, and a certain cost weight is given to each edge, so that services are more inclined to be placed on those edge server nodes with low load, rather than on some edge nodes with low load, but no It will increase too much theoretical access cost, so as to achieve load balancing of edge server nodes, and the degree of balance can be ensured by adjusting the cost weight on the edge from the edge server node to the endpoint, which can flexibly meet the The needs of different users, and can deal with many different scenarios.

Step S4: Solve the model to obtain the minimum cost and maximum flow, and place the service node on the corresponding edge server node to complete the migration and placement of the service node.

Specifically, the classical minimum cost and maximum flow solution algorithm is used to solve the problem. In the solved minimum cost and maximum flow, if there is a flow from the service node Ni through the edge server node Ej, it means that the service node Ni is placed on the edge. On the server node Ej, the service node is migrated and placed, that is, the position of the edge server node corresponding to the service node can be obtained according to the solution result, and then the service node is migrated and placed. In the embodiment of the present invention, the classical minimum cost maximum flow solution algorithm adopts continuous shortest path method, Cost Scaling, scaling method or elimination circle method, all of which can reliably and effectively obtain the minimum cost maximum flow of the model.

Step S5: Update the access information for each service, that is, re-statistic the access information for each service every time period, adjust the weight of the edge connecting the edge server node to the endpoint, and update the service node at the edge. The location on the server node to achieve load balancing, as shown in Figure 4, specifically includes:

Step S51: in the residual network of the model solving algorithm, update the access cost and migration cost of each service node placed on the edge server node;

Step S52: according to the updated access information, perform corresponding deletions and additions to the service nodes and edge server nodes in the residual network;

Step S53: Adjust the weight of the edge connecting the edge server node to the end node according to the current edge network index change and the load of each edge server node, so that the calculated placement scheme has a certain tendency, but Doesn't add much to the total access cost.

Step S54: Use the incremental minimum cost and maximum flow algorithm to quickly solve the problem, migrate and place the service node according to the obtained minimum cost and maximum flow, and update its position on the edge server node.

In the embodiment of the present invention, since the access mode of the user to the service, the resource usage of the edge node, and the network traffic will change from time to time, the access information of the service may be re-statistical at intervals. According to the previous method, the process of steps S3 and S4 will be chosen to be re-executed, but the solution time is often too long. The embodiment of the present invention adopts the incremental minimum cost maximum flow algorithm, and directly modifies the corresponding service node and the edge connected to the edge server node in the last residual network, instead of constructing the minimum cost maximum flow model from scratch and Solving, performing the deloop method on the modified residual network reduces the solving time by a factor of 8-15.

Step S6: Perform real-time monitoring and statistics on the access information of each service, and adjust the corresponding service node and edge server node when the change rate of the access information exceeds the set threshold.

Specifically, the number of user's access to the service node will have a sudden increase in volume and abnormal situation, and the access information will be monitored and counted in real time through the monitor. An optimal threshold is trained by the reinforcement learning method. When the change rate of the monitored or statistical access information exceeds the set threshold, that is, the optimal threshold, the residual network is immediately updated, and the edge server nodes are adjusted in time to connect to The weight of the edge of the termination point can be solved by the incremental minimum cost maximum flow algorithm proposed in step S5, and the service node and the edge server node are adjusted accordingly.

In the embodiment of the present invention, each service node is updated by using the incremental minimum cost maximum flow algorithm, which specifically includes the following:

Step S541: The user's access to the service node changes, and when the set conditions are met, the location of the service node is updated, specifically including:

Step S5411: When the variable of the number of visits is greater than the pre-learned threshold, in the last residual network of the solution algorithm, modify the weights of the edges connected to the service node; otherwise, do not modify the weights that need to be modified. the number of sides.

In the embodiment of the present invention, it is considered that the amount of visits to most service nodes will change every period of time, and since the number of visits by users to some service nodes will not change too much, the number of visits can be Compare with the optimal threshold learned by machine learning method in advance to determine whether to modify the weight of the edge. Further, the machine learning method is used to determine the threshold value that needs to be modified for the associated edge weights of the service nodes. Different from the traditional method, the optimal threshold value is obtained through the machine learning method and the previous data set training. The threshold obtained at this time can adapt to various access modes of users, large-scale and high-frequency changing network structure, and various abnormal situations, so that the decision center will not frequently change the residual network structure, thereby reducing the number of service adjustments and improving service quality.

Step S5412: Add a migration cost on the edge connecting the service node to the edge server node, and start from the service node whose cost has changed, and re-find the flow with the minimum cost and the maximum cost. Specifically, using the circle elimination method to find the minimum cost and maximum flow again, because it is based on the minimum cost and maximum flow in the previous stage, the circle elimination method can be used to find the optimal solution faster than other schemes.

Step S513: Update the position of the service node on the edge server node according to the obtained minimum cost and maximum flow.

In the embodiment of the present invention, it is assumed that the last partial residual network of the solution algorithm is shown in Figure 4. At this time, the capacity and weight of the edge from the service node N1 to the edge server node E1 are (0, 100), indicating that the service node has been The point N1 is placed on the edge server node E1, and the corresponding reverse edge (1, -100) is generated. Similarly, the service node Nn is allocated to the edge server node E2. After a period of time, the number of user visits to service nodes N1 and Nn changes and the number of visits is greater than the threshold, then the corresponding edges in the residual network need to be modified. The modified residual network is shown in Figure 5.

It can be seen from Figure 4 that a negative ring of N1→E2→Nn→E1→N1 is formed, and the circle elimination method can be used to eliminate the circle from the service node N1 until there is no negative circle in the residual network. At this time, the maximum flow rate is reached. Minimum fee.

Step S542: when the number of service nodes changes, delete service nodes or add service nodes, and update the locations of service nodes, specifically including:

When there are fewer service nodes, delete the service nodes and related edges in the residual network;

When a service node is added, the service node and related edges are added to the residual network, the minimum cost path from the service node to the edge server node is selected, and the flow is added to reach the maximum flow, thereby realizing all services. Node placement; add migration costs on the edge where service nodes connect to edge server nodes;

Starting from the deleted or added service node, and flowing to the edge server node, a new minimum cost maximum flow is found, and the service node is migrated and placed. Specifically, the elimination circle method is also used to find a new minimum cost maximum flow.

Step S543: When the bearing capacity of the edge server node changes, update the location of the service node, which specifically includes:

Step S5431: When the capacity of the edge server node is reduced, check whether the flow from the edge server node is larger than the capacity of the corresponding edge after the reduction, specifically:

If the outgoing flow is still less than the reduced capacity of the edge, only the capacity of the edge in the residual network will be changed, and the residual network will not be adjusted;

If the outgoing flow is larger than the reduced capacity of the edge or the edge server node is deleted, the flow corresponding to the edge with the highest cost in the edge server node will be returned; starting from the corresponding edge service node, select to end The shortest path to the point and increase the flow, and add the migration cost on the edge connecting the service node to the edge server node;

Starting from the corresponding service node, use the circle elimination method to find a new minimum cost and maximum flow, and use this to migrate and place the service node, and update the location of the service node;

Step S5432: When the capacity of the edge server node increases or a new edge server node is added, the residual network is modified, and the migration cost is added on the edge connecting the service node to the edge server node; Starting from the server node, use the circle elimination method to find a new minimum cost and maximum flow, and adjust the service node.

The multi-service migration method provided by the embodiment of the present invention establishes a minimum cost maximum flow model by constructing a network topology map and generating a shortest path distance matrix. Solve to get the minimum cost and maximum flow, migrate and place the service node, and update the position of the service node every time period. Real-time monitoring and statistics are performed on the access information of each service, and the corresponding service nodes and edge nodes are adjusted when necessary. The embodiment of the present invention can achieve a more balanced placement of virtual services on edge nodes under the condition that only a small part of the access cost is increased, and the balance degree can be adjusted, which is positively correlated with the increased access cost, and can be It is very flexibly adjusted according to the actual needs of users, and can also be adjusted according to the load of each edge node, and can well cope with various situations in large-scale dynamic networks.

Referring to FIG. 7 , an embodiment of the present invention further provides a load balancing multi-service migration system based on minimum cost and maximum flow. The system includes the following:

Topology map building module: It is used to build a network topology map based on the geographical location and connection relationship of edge servers.

Matrix generation module: used to calculate the shortest path distance between each pair of edge servers in the network topology diagram, and generate a shortest path distance matrix.

Model building module: used to establish a minimum-cost maximum flow model according to the distance matrix, combined with the access information of each service, the load capacity of each edge server and its relationship with the service delay, and the model includes the source node , service nodes, edge server nodes and edge nodes.

Service node migration module: used to solve the model to obtain the minimum cost and maximum flow, migrate the service nodes, and place them on the corresponding edge server nodes.

Service node update module: used to update the access information of each service, adjust the weight of the edge connecting the edge server node to the end node, and update the position of the service node on the edge server node.

Further, referring to Fig. 8, the service node update module includes:

Cost update sub-module: used to update the access cost and migration cost of each service node placed on the edge server node in the residual network of the model solving algorithm.

Node deletion and addition sub-module: It is used to perform corresponding deletion and addition of service nodes and edge server nodes in the residual network according to the updated access information.

Weight adjustment sub-module: Adjust the weight of the edge connecting the edge server node to the endpoint according to the current edge network index changes and the load of each edge server node.

Specifically, the system provided in the embodiment of the present invention is specifically used to execute the foregoing method embodiment, which is not repeated in this embodiment of the present invention.

FIG. 9 is a schematic diagram of the physical structure of an electronic device according to an embodiment of the present invention. As shown in FIG. 9 , the electronic device may include: a processor (processor) 301, a communications interface (Communications Interface) 302, and a memory (memory) 303 and a communication bus 304 , wherein the processor 301 , the communication interface 302 , and the memory 303 communicate with each other through the communication bus 304 . The processor 301 can call a computer program stored in the memory 303 and run on the processor 301 to execute the methods provided by the above embodiments, for example, including:

Update the access information for each service, adjust the weight of the edge connecting the edge server node to the endpoint, and update the position of the service node on the edge server node;

In addition, the above-mentioned logic instructions in the memory 303 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solutions of the embodiments of the present invention are essentially, or the parts that make contributions to the prior art or the parts of the technical solutions can be embodied in the form of software products, and the computer software products are stored in a storage medium , including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .

Embodiments of the present invention further provide a non-transitory computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it is implemented to perform the methods provided by the foregoing embodiments, for example, including:

The device embodiments described above are only illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in One place, or it can be distributed over multiple network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

From the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and certainly can also be implemented by hardware. Based on this understanding, the above-mentioned technical solutions can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products can be stored in computer-readable storage media, such as ROM/RAM, magnetic A disc, an optical disc, etc., includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform the methods described in various embodiments or some parts of the embodiments.

The method and system for multi-service migration based on load balancing with minimum cost and maximum flow provided by the embodiments of the present invention can obtain an optimal placement and migration strategy that minimizes the overall service access cost, and can also achieve a small increase in access when only a small part of the service access cost is minimized. Under the condition of cost, the placement of services on edge nodes is more balanced, and the degree of balance can be adjusted. The time complexity is much lower than the previous method, and the strategy can be solved in a relatively short time, which is a good response. The sudden increase in the number of service accesses and the abnormal situation.

The above-mentioned embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited by the above-mentioned embodiments, and any other changes, modifications, substitutions, combinations, The simplification should be equivalent replacement manners, which are all included in the protection scope of the present invention.

Claims

A multi-service migration method based on load balancing with minimum cost and maximum flow, characterized in that the specific steps of the method include the following:

Build a network topology map based on the geographic location and connection relationship of edge servers;

Calculate the shortest path distance between each pair of edge servers and generate a shortest path distance matrix;

According to the distance matrix, combined with the access information of each service, the carrying capacity of each edge server and its relationship with the service delay, a minimum cost maximum flow model is established, and the model includes source nodes, service nodes, edge server nodes and endpoints;

Solve the model to obtain the minimum cost and maximum flow, and place the service node on the corresponding edge server node;

Update the access information of each service, adjust the weight of the edge connecting the edge server node to the endpoint, and update the position of the service node on the edge server node to achieve load balancing;

Real-time monitoring and statistics are performed on the access information of each service. When the change rate of the access information exceeds the set threshold, the corresponding service nodes and edge server nodes are adjusted.
The multi-service migration method based on load balancing of minimum cost and maximum flow according to claim 1, wherein the minimum cost and maximum flow model is specifically:

Taking the S node and the T node as the source node and the end node of the model, Ni represents the ith service node, Ej represents the jth edge server node, the source node S is only connected to the service node Ni, each Each service node Ni is connected to all edge server nodes Ej, and each edge server node Ej is connected to the termination node T; each edge in the model is associated with capacity and communication cost.
The multi-service migration method based on load balancing with minimum cost and maximum flow according to claim 2, wherein each edge server node Ej passes through at least two edges and a termination point according to the number of services it can bear T are connected, and the cost weights of each edge are different and have an increasing relationship.
The multi-service migration method based on load balancing with minimum cost and maximum flow according to claim 2, characterized in that: the specific process of updating the position of the service node on the edge server node includes:

In the residual network of the model solving algorithm, update the access cost and migration cost of each service node placed on the edge server node;

According to the updated access information, delete and add corresponding service nodes and edge server nodes in the residual network;

Adjust the weight of the edge connecting the edge server node to the endpoint according to the current edge network index changes and the load of each edge server node;

The incremental minimum cost and maximum flow algorithm is used to solve the problem, and the position of the service node on the edge server node is updated according to the obtained minimum cost and maximum flow.
The multi-service migration method based on load balancing with minimum cost and maximum flow according to claim 4, wherein the incremental minimum cost and maximum flow algorithm specifically includes the following:

The access volume of the service node changes, and when the variable of access times is greater than the pre-learned threshold, the location of the service node is updated;

When the number of service nodes changes, delete service nodes or add service nodes, and update the location of service nodes;

When the capacity of the edge server node changes, and when the capacity of the edge server node decreases and the flow from it is larger than the reduced capacity of the edge, or when the edge server node is deleted, and when the capacity of the edge server node increases Or when a new edge server node is added, the location of the service node is updated.
The load balancing multi-service migration method based on the minimum cost and maximum flow according to claim 5, characterized in that: starting from the changed node, re-find a new minimum cost and maximum flow, and carry out the process for the service node. Migration and placement, updating the location of service nodes on edge server nodes.
A load balancing multi-service migration system based on minimum cost and maximum flow, characterized in that: the system includes the following:

Topology map building module: used to build a network topology map according to the geographical location and connection relationship of edge servers;

Matrix generation module: used to calculate the shortest path distance between each pair of edge servers in the network topology diagram, and generate a shortest path distance matrix;

Model building module: used to establish a minimum-cost maximum flow model according to the distance matrix, combined with the access information of each service, the load capacity of each edge server and its relationship with the service delay, and the model includes the source node , service nodes, edge server nodes and edge nodes;

Service node migration module: used to solve the model to obtain the minimum cost and maximum flow, migrate the service nodes, and place them on the corresponding edge server nodes;

Service node update module: used to update the access information for each service, adjust the weight of the edge connecting the edge server node to the endpoint, and update the position of the service node on the edge server node;

Monitoring and adjustment module: It is used to monitor and count the access information of each service in real time, and adjust the corresponding service nodes and edge server nodes when the change rate of the access information exceeds the set threshold.
The load balancing multi-service migration system based on the minimum cost and maximum flow according to claim 7, wherein the service node update module comprises:

Cost update sub-module: used to update the access cost and migration cost of each service node placed on the edge server node in the residual network of the model solving algorithm;

Node deletion and addition sub-module: It is used to delete and add service nodes and edge server nodes in the residual network according to the updated access information;

Weight adjustment sub-module: adjust the weight of the edge connecting the edge server node to the endpoint according to the current edge network index changes and the load of each edge server node;

Node update sub-module: use the incremental minimum cost and maximum flow algorithm to solve, and update the position of the service node on the edge server node according to the minimum cost and maximum flow obtained by the solution.
An electronic device, comprising a memory, a processor and a computer program stored on the memory and running on the processor, characterized in that, when the processor executes the program, the program as claimed in any one of claims 1 to 7 is implemented. steps of the method described.
A non-transitory computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.