CN113132497A - Load balancing and scheduling method for mobile edge operation - Google Patents

Load balancing and scheduling method for mobile edge operation Download PDF

Info

Publication number
CN113132497A
CN113132497A CN202110676139.8A CN202110676139A CN113132497A CN 113132497 A CN113132497 A CN 113132497A CN 202110676139 A CN202110676139 A CN 202110676139A CN 113132497 A CN113132497 A CN 113132497A
Authority
CN
China
Prior art keywords
user equipment
cluster
edge
edge node
node
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110676139.8A
Other languages
Chinese (zh)
Other versions
CN113132497B (en
Inventor
傅志愿
张康崇
聂世元
叶颖哲
鲍其炜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Tian Ship Information Technology Ltd By Share Ltd
Original Assignee
Hangzhou Tian Ship Information Technology Ltd By Share Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Tian Ship Information Technology Ltd By Share Ltd filed Critical Hangzhou Tian Ship Information Technology Ltd By Share Ltd
Priority to CN202110676139.8A priority Critical patent/CN113132497B/en
Publication of CN113132497A publication Critical patent/CN113132497A/en
Application granted granted Critical
Publication of CN113132497B publication Critical patent/CN113132497B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
    • H04L67/1004—Server selection for load balancing
    • H04L67/1021—Server selection for load balancing based on client or server locations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5061—Partitioning or combining of resources
    • G06F9/5072—Grid computing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5083—Techniques for rebalancing the load in a distributed system
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
    • H04L67/1004—Server selection for load balancing
    • H04L67/1008—Server selection for load balancing based on parameters of servers, e.g. available memory or workload
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/0226—Traffic management, e.g. flow control or congestion control based on location or mobility
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/08—Load balancing or load distribution

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Hardware Design (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

A load balancing and scheduling method for mobile edge operation belongs to the technical field of flow control and comprises the following steps: step S1, clustering the user equipment by using a K-means algorithm to obtain initial cluster deployment information of the user equipment; step S2, the cloud verifies whether the preliminary cluster deployment information of the user equipment meets load balancing; step S3, when the single ue is displaced, the load scheduling mechanism is triggered. According to the scheme, the K-means algorithm is used for clustering the user equipment, and then the service copy of the user equipment is deployed to the edge server closest to the clustering center, so that the distance between the user equipment and the edge node is reduced, and the response time of the service is reduced.

Description

Load balancing and scheduling method for mobile edge operation
Technical Field
The invention belongs to the technical field of flow control, and particularly relates to a load balancing and scheduling method for mobile edge operation.
Background
Mobile Edge Computing (MEC) is a concept of network architecture, and provides cloud computing capability at the edge of a mobile network and an IT service environment. Compared with the traditional network architecture and mode, the MEC has many obvious advantages, can solve the problems of high time delay, low efficiency and the like in the traditional network architecture and mode, and is just the advantages, so that the MEC becomes a key technology of 5G.
The MEC sinks the calculation and storage capacity to the edge of the network, and because the MEC is closer to the user, the user request does not need to pass through a long transmission network to reach a remote core network to be processed any more, but a part of traffic is unloaded by a locally deployed MEC server, and is directly processed and responded to the user, the communication delay is greatly reduced. Taking video transmission as an example, in a traditional mode without using MEC, when each user terminal initiates a video content call request, the user terminal first needs to access through a base station, then connects with target content through a core network, and then returns layer by layer, and finally completes interaction between the terminal and the target content, such a connection and layer-by-layer acquisition mode is very time-consuming. After introducing the MEC solution, an MEC server is deployed at a base station side close to a user, the content is cached on the MEC server by using a storage resource provided by the MEC, the user can directly obtain the content from the MEC server, and the user does not need to obtain the content data from a relatively far core network through a long backhaul link. Therefore, the waiting time between the request sending and the response sending of the user can be greatly saved, and the service quality experience of the user is improved.
However, the computing resources of the MEC server are limited, and if too many users request services from the same edge server, the bandwidth will be reduced. In addition, if the service deployment is unbalanced, some edge servers may be idle for a long time, thereby causing waste of resources.
Chinese patent publication No. CN112601256A discloses a load scheduling method based on MEC-SBS clustering in an ultra-dense network, which, as shown in fig. 1, partitions MEC-SBS in the system into a plurality of non-overlapping calculation cooperation clusters by using a partition algorithm, thereby implementing a conversion of a large-scale MEC-SBS calculation cooperation problem into a small-scale MEC-SBS calculation cooperation problem in the calculation cooperation clusters. However, it has the following disadvantages:
1, the scheme adopts a k-means clustering algorithm to construct an initial cooperative cluster, but does not consider the state of an edge server and the number of service people. If the number of service people is too large, the bandwidth is reduced, and the network communication quality is deteriorated.
2, the mobile ue is dynamic, and the initial clustering is not necessarily suitable for the ue after subsequent moving, so the mobile ue needs to switch the edge server connected to it. For the conversion mode, the chinese patent of CN112601256A discloses step two, offloading the calculation task; the mobile user equipment selects the MEC-SBS with the best channel gain to associate with, and then offloads its resulting computational tasks to its associated MEC-SBS. However, in this step, only for the initialized adjustment, not for the adjustment of the single ue due to the position movement, the switching mechanism of the single ue is lacking.
Disclosure of Invention
In view of the above-mentioned shortcomings in the prior art, the present invention is directed to a method for load balancing and scheduling of mobile edge operations.
In order to achieve the above object, the present invention adopts the following technical solutions.
A load balancing and scheduling method for mobile edge operation comprises the following steps:
step S1, clustering the user equipment by using a K-means algorithm to obtain initial cluster deployment information of the user equipment;
step S1a, the user equipment selects one edge node as the main node, all the other edge nodes are the slave nodes, and sends the selection information to the main node;
step S1b, the main node sets the longitude and latitude coordinates of all edge nodes as the initial central point of the cluster; the main node informs the rest edge nodes as slave nodes and collects longitude and latitude coordinate information of all the edge nodes;
step S1c, the slave node calculates the user device connected to it to the nearest user device
The distance of the initial central point of the cluster is obtained, the user equipment closest to the initial central point of the same cluster belongs to one cluster, and then the distance information and the cluster information are transmitted back to the main node;
step S1d, after receiving the distance information and cluster information returned by all the slave nodes, the master node recalculates the new central point of each cluster; selecting the cluster number of the cluster as the number k of edge nodes, and then judging whether a convergence condition is reached;
if the convergence condition is not reached, returning to step S1b to reset the cluster center point: according to the user equipment in each cluster, calculating the central points of the user equipment, and taking the central points as new central points of the cluster; repeating the step S1b to the step S1d to continue iterative operation until the calculation result reaches a convergence condition;
if the convergence condition is reached, the main node corresponds the last cluster center point to the actual edge node closest to the main node;
step S1e, outputting a cluster set to obtain initial cluster deployment information of user equipment, and sending the information to a cloud; each user equipment is uniquely attributed to only one cluster;
step S2, the cloud verifies whether the preliminary cluster deployment information of the user equipment meets load balancing;
step S3, when the single ue is displaced, the load scheduling mechanism is triggered.
Further, in step S1d, the convergence condition is set to be that all the cluster center point movements are smaller than the threshold, i.e. the clustering criterion function
Figure DEST_PATH_IMAGE001
Not more than a threshold value; in the aggregation criterion function J, k is the number of central points, i represents the serial number of the cluster, i is more than or equal to 1 and less than or equal to k,xare clustersC iThe location of the user equipment of (a),μ iare clustersC iThe position of the center point.
3. The method for load balancing and scheduling of mobile edge operations as claimed in claim 2, wherein the step S2 comprises the steps of:
step S2a, aiming at each edge node in the edge node set E, sequentially checking whether the condition of excess load occurs; wherein a set of edge nodes E = { E = { (E)0,E1…EkK is the number of edge nodes;
if Ii>LiThen, it represents the ith edge node EiIf the load is exceeded, triggering a load balancing mechanism, and turning to the step S2 b; wherein, IiDenotes the ith edge node EiThe number of the user equipment allocated at present; l isiDenotes the ith edge node EiThe maximum number of user equipment can be borne;
otherwise, go to step S3;
step S2b, edge node E of overloadiNumber of currently allocated user equipments IiSubtracting the maximum user equipment quantity L borne by the edge nodeiObtaining the number O of the user equipment with the edge node exceeding the loadi；
Step S2c, according to the preliminary cluster deployment information of the user equipment and the deployment information of the edge nodes, the edge nodes which do not exceed the load are compared with the edge nodes E according to the deployment information of the edge nodesiIs sorted from small to large and the nearest edge node E is selectedtTo share EiExcess number of user devices;
step S2d, EtThe maximum number of user equipments L that can be bornetSubtract EtNumber of currently allocated user equipments ItObtaining EtAffordable number of user equipments Rt；
Step S2e, if Oi>RtThen represent EtCannot all accept EiThe number of the user equipment exceeding the load can only accept R at mosttA user equipment and starts to adjust EtNumber of user equipments deployed, will EtAccepted RtA user equipment and EtA communication connection; after the adjustment is completed, the number O of the overload user equipment which is not distributed is recalculatediThen returning to the step S2c, continuing iterative operation to find the edge node of the next acceptable overloaded user equipment;
if O isi<RtThen represent EtCan all accept EiAll overloaded user equipments and start to adjust EtNumber of user equipments deployed, will EtAccepted OiA user equipment and EtA communication connection; after the adjustment is finished, finishing the iterative operation to complete EiAdjusting the load;
step S2f, sequentially and continuously checking each edge node in the E set, and repeating the step S2b to the step S2E until all edge nodes do not exceed the load;
and step S2g, the cloud sends information to inform the edge node of the service transfer process, the edge node sends information to inform the user equipment connected with the edge node of changing the connection line, and the user equipment belonging to the same cluster is moved to the edge server corresponding to the cluster center point.
Further, step S3 includes the following steps:
step S3a, the user equipment periodically transmits heartbeat information to the edge node in communication connection with the user equipment, where the heartbeat information includes: identification number, latitude, longitude, base station name and local area network address of current connection and heartbeat sending time of the user device;
step S3b, the edge node actively detects whether the user equipment is displaced; the edge node judges whether the user equipment moves according to the local area network address position of the user equipment connecting with the base station: when the user equipment moves, the connected base station is switched, the local area network address of the connected base station is changed, which indicates that the user equipment has shifted, and the moving direction of the user equipment can be known through the switching of the local area network of the base station; when the user equipment converts the displacement, it is indicated that a more suitable edge node can provide service for the user equipment, and the step S3c is carried out;
step S3c, the edge node corresponding to the base station of the user equipment switching connection is used as the target edge node, and the information of the target edge node is sent to the user equipment to be scheduled;
step S3d, after receiving the information of the target edge node, the user equipment transfers the system channel to connect to the target edge node, then records the service related information and converts the service channel; and after the conversion is finished, the user equipment is transferred to the target edge node to continue using the service, and the load scheduling mechanism is executed.
The invention can achieve the following technical effects:
1. according to the scheme, the K-means algorithm is used for clustering the user equipment, and then the service copy of the user equipment is deployed to the edge server closest to the clustering center, so that the distance between the user equipment and the edge node is reduced, and the response time of the service is reduced.
2. The scheme adopts a load balancing mechanism, and overcomes the defect that the load balancing of the edge server is not considered by the K-means algorithm. The scheme transmits the calculated deployment result back to the cloud, and the cloud verifies whether any edge node is overloaded: if so, starting a load balancing mechanism, adjusting a deployment result according to the load capacity of the edge node, and performing iterative operation to find the next edge node until the load balancing mechanism is executed. According to the scheme, the situation that the bandwidth is reduced and the network response time is further increased due to the fact that too many users request services from the same edge server is avoided, and the number of the edge server service people is adjusted according to the information such as the state, the capacity and the service deployment position of the edge server.
3. The scheme adopts a load scheduling mechanism and a heartbeat mechanism, monitors the moving position of the user equipment, grasps the state of the user equipment in real time, and adjusts the edge server connected with the user equipment in due time when the user equipment moves, so that the user equipment can also keep certain service quality in the moving process.
4. When the mobile user equipment moves to a network domain where no edge server is deployed, the primary edge server still requires a suitable edge server to provide a connection for the mobile user equipment according to the location of the mobile user equipment. Although the mobile user equipment and the edge server are not in the same network domain, the connection mode still has the advantage of lower network delay time compared with a remote cloud center.
Experimental results show that the scheme can effectively determine the appropriate deployment position of the edge node, and effectively reduce 5-22% of response time of service required by the user equipment.
Drawings
FIG. 1 is an exemplary diagram of a wireless sensor network structure of the Chinese patent application with publication number CN 101478826A;
FIG. 2 is a flow chart of a conventional k-means clustering algorithm;
FIG. 3 is a flowchart of steps S1 and S2 of the present invention;
FIG. 4 is a flowchart of step S3 of the present invention;
fig. 5 is a diagram illustrating an example of a network structure of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
Mobile Edge Computing (MEC) is an evolution of cloud computing, sinks application hosting from a centralized data center to the edge of a network, is closer to data generated by consumers and applications, provides IT and cloud computing capabilities near the edge of the network of mobile users, and opens up opportunities for obtaining high bandwidth, low latency, and near-end deployment advantages by using network capabilities, thereby creating new services and revenue and creating new business models.
The MEC is one of the key technologies to realize 5G low latency and boost bandwidth rate, etc., while the MEC opens up a network edge for applications and services, including applications and services from third parties, so that the communication network can be transformed into a multifunctional service platform for other industries and specific customer groups.
The MEC basically adopts a three-layer architecture, from near to far, which is a mobile user equipment, an edge node and a cloud.
The cloud is used for processing larger tasks and continuous services.
The edge node is deployed near the base station and is responsible for assisting the cloud in deploying services, collecting data and mastering the state of the mobile user equipment.
The basic idea of the k-means cluster algorithm is that for a given sample set, the sample set is divided into k clusters according to the distance between samples, points in the clusters are connected together as closely as possible, and the distance between the clusters is as large as possible. The method comprises the following basic steps:
step 1: the number k of clusters to be clustered is selected, and k center points are selected.
step 2: for each sample point, the closest central point (finding tissue) is found, and the points closest to the same central point are a class, thus completing one clustering.
step 3: and judging whether the class conditions of the sample points before and after the cluster are the same, if so, terminating the algorithm, and otherwise, entering step 4.
step 4: for the sample points in each category, the center points of these sample points are calculated, and step2 is continued as the new center point for that category.
As shown in FIG. 2, the K-Means algorithm flows as follows:
1) input is a sample set D = ∑ tonex 1,x 2,...x m}，（x iFor the ith sample, m is the total number of samples), the cluster number k of the cluster. Randomly select k samples from the sample set D as the initial k center point vectors: { mu. }1,μ2,...,μkTherein ofμ iIs the ith central point;
2) calculating the distance from each sample point to each central point and selecting the cluster closest to the data point as the cluster; recalculating the center point of each cluster according to the result of each sample point selection; then judging whether a convergence condition is reached:
a) initializing cluster partitioning C to Ct=Φ，（t=1,2...k）；
b) For i =1,2.. m, calculate samplesx iAnd respective center pointsμ jDistance of (j =1,2.. k): dij=||x i−μ j||2 2Will bex iMinimum mark is dijCorresponding class λi. At this time, update Cλi=Cλi∪{x i}；
c) For j =1,2jAll sample points in the image recalculate a new center point
Figure 223522DEST_PATH_IMAGE002
；
e) The convergence condition is the moving amplitude of the central point or the set iteration round number; the moving amplitude of the central point is less than the set value, and then the cluster criterion function is performed
Figure 609504DEST_PATH_IMAGE003
Converging and turning to the step 3); otherwise, turning to the step b) to continue iteration until the moving amplitude of the central point is smaller than a set value or the number of iteration rounds reaches the set value;
in the convergence criterion function J, k is the number of the central points,xare clustersC iThe number of the sample points of (a),μ iis the ith center point.
3) Output cluster C = { C = { C =1,C2,...Ck}。
The K-means algorithm generally uses the clustering criterion function J as an objective function, and the objective is to find the minimum value of the clustering criterion function. The sample points can be divided into K clusters by the K-means algorithm. The advantage of using the K-means algorithm is that the user equipment can be deployed near the nearest edge server, thereby effectively reducing the response time of the whole network.
However, the K-means algorithm does not take the load of the edge server into consideration, and when the sample points (user devices) are too concentrated, the central point (edge server) at the dense sample points will run overloaded, so it is necessary to introduce a load balancing mechanism based on the K-means algorithm.
Meanwhile, the K-means algorithm initializes the layout of the ue, but does not consider that the mobile ue is dynamic, and the initial clustering is not necessarily suitable for the ue after subsequent mobility. Therefore, it is necessary to introduce a load scheduling mechanism based on the K-means algorithm, and the deployment can be adjusted in real time according to the current state of the user equipment.
A method for load balancing and scheduling of mobile edge operations, as shown in fig. 3, 4 and 5, comprises the following steps:
and step S1, clustering the user equipment by using a K-means algorithm to obtain initial cluster deployment information of the user equipment.
Step S1a, the user equipment selects one edge node as the main node, all the other edge nodes are the slave nodes, and sends the selection information to the main node;
step S1b, the main node sets the longitude and latitude coordinates of all edge nodes as the initial central point of the cluster; and the master node informs the rest edge nodes of being slave nodes and collects longitude and latitude coordinate information of all the edge nodes.
Step S1c, the slave node calculates the distance from the user equipment in communication connection with the slave node to the initial center point of the cluster closest to the user equipment, assigns the user equipment closest to the initial center point of the same cluster to a cluster, and then transmits the distance information and cluster information back to the master node.
Step S1d, after receiving the distance information and cluster information returned by all the slave nodes, the master node recalculates the new central point of each cluster; and selecting the cluster number of the cluster as the number k of the edge nodes, and then judging whether a convergence condition is reached.
The convergence criterion is set to be less than a threshold value, e.g., less than one meter, for all cluster center point movements, i.e., the clustering criterion function
Figure 982717DEST_PATH_IMAGE004
Less than or equal to 1; in the aggregation criterion function J, k is the number of center points (also the number of edge nodes), i represents the serial number of the cluster, i is more than or equal to 1 and less than or equal to k,xare clustersC iThe location of the user equipment of (a),μ iare clustersC iThe position of the center point.
If the convergence condition is not reached, returning to step S1b to reset the cluster center point: according to the user equipment in each cluster, calculating the central points of the user equipment, and taking the central points as new central points of the cluster; the iterative operations from step S1b to step S1d are repeated until the calculation result (convergence criterion function J) reaches the convergence condition.
If the convergence condition is reached, the master node corresponds the last cluster center point to the actual edge node closest to it.
Step S1e, outputting the cluster set, obtaining the preliminary cluster deployment information of the user equipment, and sending the information to the cloud. Each user equipment is uniquely attributed to only one cluster.
Since the center point of the cluster does not have to have the edge node of the exact entity, the center point of the cluster is corresponding to the edge node closest to the center point of the cluster, and it is possible that a plurality of user equipments may correspond to the same edge node, resulting in unbalanced load of the edge node.
Step S2, the cloud verifies whether the preliminary cluster deployment information of the user equipment satisfies load balancing.
Step S2a, aiming at each edge node in the edge node set E, sequentially checking whether the condition of excess load occurs; wherein a set of edge nodes E = { E = { (E)0,E1…EkAnd k is the number of edge nodes.
If Ii>LiThen, it represents the ith edge node EiIf the load is exceeded, triggering a load balancing mechanism, and turning to the step S2 b; wherein, IiDenotes the ith edge node EiThe number of the user equipment allocated at present; l isiDenotes the ith edge node EiThe maximum number of user equipment can be borne;
otherwise, the process proceeds to step S3.
Step S2b, edge node E of overloadiNumber of currently allocated user equipments IiSubtracting the maximum user equipment quantity L borne by the edge nodeiObtaining the number O of the user equipment with the edge node exceeding the loadi。
Step S2c, according to the preliminary cluster deployment information of the user equipment and the deployment information of the edge nodes, the edge nodes which do not exceed the load are compared with the edge nodes E according to the deployment information of the edge nodesiIs sorted from small to large and the nearest edge node E is selectedtTo share EiToo many user equipments.
Step S2d, EtThe maximum number of user equipments L that can be bornetSubtract EtNumber of currently allocated user equipments ItObtaining EtAffordable number of user equipments Rt。
Step S2e, if Oi>RtThen represent EtCannot all accept EiThe number of the user equipment exceeding the load can only accept R at mosttA user equipment and starts to adjust EtNumber of user equipments deployed, will EtAccepted RtA user equipment and EtA communication connection; after the adjustment is completed, the number O of the overload user equipment which is not distributed is recalculatediThen, returning to step S2c, the iterative operation is continued to find the edge node of the next acceptable overloaded user equipment.
If O isi<RtThen represent EtCan all accept EiAll overloaded user equipments and start to adjust EtNumber of user equipments deployed, will EtAccepted OiA user equipment and EtA communication connection; after the adjustment is finished, finishing the iterative operation to complete EiAnd (4) adjusting the load.
Step S2f, sequentially checking each edge node in the E set, and repeating steps S2b to S2E until all edge nodes do not exceed the load.
And step S2g, the cloud sends information to inform the edge node of the service transfer process, the edge node sends information to inform the user equipment connected with the edge node of changing the connection line, and the user equipment belonging to the same cluster is moved to the edge server corresponding to the cluster center point.
According to the scheme, after the K-means calculation is finished, the load evaluation of the edge nodes is carried out according to the calculation result, the cloud end can master the load capacity of all the edge nodes, therefore, a centralized load mechanism of the cloud end is adopted, when the cloud end confirms that the deployment of all the services is within the range which can be borne by the edge nodes, the service user equipment transfer message is sent, and the user equipment is deployed at a proper position.
Step S3, when the single ue is displaced, the load scheduling mechanism is triggered.
Step S3a, the user equipment periodically transmits heartbeat information to the edge node in communication connection with the user equipment, where the heartbeat information includes: the identification number, latitude, longitude, base station name and local area network address of the current connection line and heartbeat sending time of the user device.
In step S3b, the edge node actively detects whether the ue is moving. The edge node judges whether the user equipment moves according to the local area network address position of the user equipment connecting with the base station: when the user equipment moves, the connected base station is switched, the local area network address of the connected base station is changed, which indicates that the user equipment has shifted, and the moving direction of the user equipment can be known through the switching of the local area network of the base station; when the ue switches the shift, it indicates that there is a more suitable edge node to serve it, and the process goes to step S3 c.
Step S3c, the edge node corresponding to the base station to which the ue switches the connection is taken as a target edge node, and the information of the target edge node is sent to the ue to be scheduled.
Step S3d, after receiving the information of the target edge node, the user equipment transfers the system channel to connect to the target edge node, then records the service related information and converts the service channel; and after the conversion is finished, the user equipment is transferred to the target edge node to continue using the service, and the load scheduling mechanism is executed.
According to the scheme, a heartbeat transmission mechanism is adopted, so that the edge node can master the information and the movement of the user equipment, and therefore the user equipment needs to transmit heartbeat information regularly. And the edge node receives the heartbeat information, records and analyzes the heartbeat information, and if the user equipment is displaced, gives corresponding information to the user equipment so as to execute a load scheduling process. Compared with the scheme, if the mobile device is displaced, the mobile device requests the cloud for service response and converts a service channel, and the response time is increased by 3-4 times.
According to the scheme, the information of the mobile user equipment is transmitted to the edge server, and the edge server grasps the movement of the mobile user equipment and adjusts service deployment and the like. Compared with a centralized management mode of transmitting the information of the user equipment to the cloud, the scheme reduces network delay. Compared with an autonomous management mode that the information of the user equipment is transmitted to the edge server or the cloud end when needed, the scheme can systematically master the movement of the mobile user equipment and adjust the requirements of the user in real time. In addition, after the information of the mobile user equipment is transmitted to the edge server, the edge server transmits the information of all the mobile user equipment to the cloud, so that the cloud can master the information of all the user equipment, and the cloud can provide various system services.
It is worth pointing out that chinese patent publication No. CN112601256A discloses a load scheduling method based on MEC-SBS clustering in an ultra-dense network, which combines each cluster to form a computation cooperation cluster (see fig. 1 thereof), and a sample point in a k-means clustering algorithm is MEC-SBS (mobile edge computation server). In the scheme, the user equipment is clustered to be connected to the most appropriate edge server, and the sample point in the k-means clustering algorithm is the user equipment. Therefore, the calculation objects of both are not the same.
It should be understood that equivalents and modifications of the technical solution and inventive concept thereof may occur to those skilled in the art, and all such modifications and alterations should fall within the scope of the appended claims.

Claims (4)

1. A load balancing and scheduling method for mobile edge operation is characterized by comprising the following steps:
step S1, clustering the user equipment by using a K-means algorithm to obtain initial cluster deployment information of the user equipment;
step S1a, the user equipment selects one edge node as the main node, all the other edge nodes are the slave nodes, and sends the selection information to the main node;
step S1b, the main node sets the longitude and latitude coordinates of all edge nodes as the initial central point of the cluster; the main node informs the rest edge nodes as slave nodes and collects longitude and latitude coordinate information of all the edge nodes;
step S1c, the slave node calculates the user device connected to it to the nearest user device
The distance of the initial central point of the cluster is obtained, the user equipment closest to the initial central point of the same cluster belongs to one cluster, and then the distance information and the cluster information are transmitted back to the main node;
step S1d, after receiving the distance information and cluster information returned by all the slave nodes, the master node recalculates the new central point of each cluster; selecting the cluster number of the cluster as the number k of edge nodes, and then judging whether a convergence condition is reached;
if the convergence condition is not reached, returning to step S1b to reset the cluster center point: according to the user equipment in each cluster, calculating the central points of the user equipment, and taking the central points as new central points of the cluster; repeating the step S1b to the step S1d to continue iterative operation until the calculation result reaches a convergence condition;
if the convergence condition is reached, the main node corresponds the last cluster center point to the actual edge node closest to the main node;
step S1e, outputting a cluster set to obtain initial cluster deployment information of user equipment, and sending the information to a cloud; each user equipment is uniquely attributed to only one cluster;
step S2, the cloud verifies whether the preliminary cluster deployment information of the user equipment meets load balancing;
step S3, when the single ue is displaced, the load scheduling mechanism is triggered.
2. The method of claim 1, wherein in step S1d, the convergence criterion is set to be that all cluster center point shifts are less than a threshold value, i.e. the clustering criterion function
Figure 925564DEST_PATH_IMAGE002
Not more than a threshold value; in the aggregation criterion function J, k is the number of central points, i represents the serial number of the cluster, i is more than or equal to 1 and less than or equal to k,xare clustersC iThe location of the user equipment of (a),μ iare clustersC iThe position of the center point.
3. The method for load balancing and scheduling of mobile edge operations as claimed in claim 2, wherein the step S2 comprises the steps of:
step S2a, aiming at each edge node in the edge node set E, sequentially checking whether the condition of excess load occurs; wherein a set of edge nodes E = { E = { (E)0,E1…EkK is the number of edge nodes;
if Ii>LiThen, it represents the ith edge node EiIf the load is exceeded, triggering a load balancing mechanism, and turning to the step S2 b; wherein, IiDenotes the ith edge node EiThe number of the user equipment allocated at present; l isiDenotes the ith edge node EiThe maximum number of user equipment can be borne;
otherwise, go to step S3;
step S2b, edge node E of overloadiNumber of currently allocated user equipments IiSubtracting the maximum user equipment quantity L borne by the edge nodeiObtaining the number O of the user equipment with the edge node exceeding the loadi；
Step S2c, according to the preliminary cluster deployment information of the user equipment and the deployment information of the edge nodes, the edge nodes which do not exceed the load are compared with the edge nodes E according to the deployment information of the edge nodesiIs sorted from small to large and the nearest edge node E is selectedtTo share EiExcess number of user devices;
step S2d, EtThe maximum number of user equipments L that can be bornetSubtract EtNumber of currently allocated user equipments ItObtaining EtAffordable number of user equipments Rt；
Step S2e, if Oi>RtThen represent EtCannot all accept EiThe number of the user equipment exceeding the load can only accept R at mosttA user equipment and starts to adjust EtNumber of user equipments deployed, will EtAccepted RtA user equipment and EtA communication connection; after the adjustment is completed, the number O of the overload user equipment which is not distributed is recalculatediThen returning to the step S2c, continuing iterative operation to find the edge node of the next acceptable overloaded user equipment;
if O isi<RtThen represent EtCan all accept EiAll overloaded user equipments and start to adjust EtNumber of user equipments deployed, will EtAccepted OiA user equipment and EtA communication connection; after the adjustment is finished, finishing the iterative operation to complete EiAdjusting the load;
step S2f, sequentially and continuously checking each edge node in the E set, and repeating the step S2b to the step S2E until all edge nodes do not exceed the load;
and step S2g, the cloud sends information to inform the edge node of the service transfer process, the edge node sends information to inform the user equipment connected with the edge node of changing the connection line, and the user equipment belonging to the same cluster is moved to the edge server corresponding to the cluster center point.
4. The method for load balancing and scheduling of mobile edge operations according to claim 1 or 3, wherein the step S3 comprises the following steps:
step S3a, the user equipment periodically transmits heartbeat information to the edge node in communication connection with the user equipment, where the heartbeat information includes: identification number, latitude, longitude, base station name and local area network address of current connection and heartbeat sending time of the user device;
step S3b, the edge node actively detects whether the user equipment is displaced; the edge node judges whether the user equipment moves according to the local area network address position of the user equipment connecting with the base station: when the user equipment moves, the connected base station is switched, the local area network address of the connected base station is changed, which indicates that the user equipment has shifted, and the moving direction of the user equipment can be known through the switching of the local area network of the base station; when the user equipment converts the displacement, it is indicated that a more suitable edge node can provide service for the user equipment, and the step S3c is carried out;
step S3c, the edge node corresponding to the base station of the user equipment switching connection is used as the target edge node, and the information of the target edge node is sent to the user equipment to be scheduled;
step S3d, after receiving the information of the target edge node, the user equipment transfers the system channel to connect to the target edge node, then records the service related information and converts the service channel; and after the conversion is finished, the user equipment is transferred to the target edge node to continue using the service, and the load scheduling mechanism is executed.
CN202110676139.8A 2021-06-18 2021-06-18 Load balancing and scheduling method for mobile edge operation Active CN113132497B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110676139.8A CN113132497B (en) 2021-06-18 2021-06-18 Load balancing and scheduling method for mobile edge operation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110676139.8A CN113132497B (en) 2021-06-18 2021-06-18 Load balancing and scheduling method for mobile edge operation

Publications (2)

Publication Number Publication Date
CN113132497A true CN113132497A (en) 2021-07-16
CN113132497B CN113132497B (en) 2021-09-10

Family

ID=76783189

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110676139.8A Active CN113132497B (en) 2021-06-18 2021-06-18 Load balancing and scheduling method for mobile edge operation

Country Status (1)

Country Link
CN (1) CN113132497B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115809148A (en) * 2023-01-16 2023-03-17 中国华能集团清洁能源技术研究院有限公司 Load balancing task scheduling method and device facing edge computing
TWI821038B (en) * 2022-11-22 2023-11-01 財團法人工業技術研究院 Computing task dispatching method, terminal electronic device and computing system using the same
TWI836450B (en) * 2021-09-27 2024-03-21 美商萬國商業機器公司 GENERATIVE ADVERSARIAL NETWORKS (GANs) BASED IDENTIFICATION OF AN EDGE SERVER

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102075359A (en) * 2011-01-17 2011-05-25 北京瑞汛世纪科技有限公司 Network coordinate-based server deployment method and device
US20190243438A1 (en) * 2018-02-08 2019-08-08 Korea Advanced Institute Of Science And Technology Method and system for deploying dynamic virtual object for reducing power in mobile edge computing environment
CN111586762A (en) * 2020-04-29 2020-08-25 重庆邮电大学 Task unloading and resource allocation joint optimization method based on edge cooperation
CN111866949A (en) * 2020-07-31 2020-10-30 西安工业大学 Method, system, equipment and storage medium for setting edge server in smart city
CN111949404A (en) * 2020-08-12 2020-11-17 北京金山云网络技术有限公司 Method and device for adjusting server load and related equipment
CN112601256A (en) * 2020-12-07 2021-04-02 广西师范大学 MEC-SBS clustering-based load scheduling method in ultra-dense network

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102075359A (en) * 2011-01-17 2011-05-25 北京瑞汛世纪科技有限公司 Network coordinate-based server deployment method and device
US20190243438A1 (en) * 2018-02-08 2019-08-08 Korea Advanced Institute Of Science And Technology Method and system for deploying dynamic virtual object for reducing power in mobile edge computing environment
CN111586762A (en) * 2020-04-29 2020-08-25 重庆邮电大学 Task unloading and resource allocation joint optimization method based on edge cooperation
CN111866949A (en) * 2020-07-31 2020-10-30 西安工业大学 Method, system, equipment and storage medium for setting edge server in smart city
CN111949404A (en) * 2020-08-12 2020-11-17 北京金山云网络技术有限公司 Method and device for adjusting server load and related equipment
CN112601256A (en) * 2020-12-07 2021-04-02 广西师范大学 MEC-SBS clustering-based load scheduling method in ultra-dense network

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI836450B (en) * 2021-09-27 2024-03-21 美商萬國商業機器公司 GENERATIVE ADVERSARIAL NETWORKS (GANs) BASED IDENTIFICATION OF AN EDGE SERVER
TWI821038B (en) * 2022-11-22 2023-11-01 財團法人工業技術研究院 Computing task dispatching method, terminal electronic device and computing system using the same
CN115809148A (en) * 2023-01-16 2023-03-17 中国华能集团清洁能源技术研究院有限公司 Load balancing task scheduling method and device facing edge computing

Also Published As

Publication number Publication date
CN113132497B (en) 2021-09-10

Similar Documents

Publication Publication Date Title
CN113132497B (en) Load balancing and scheduling method for mobile edge operation
CN110647391B (en) Edge computing method and system for satellite-ground cooperative network
CN109947545B (en) Task unloading and migration decision method based on user mobility
CN109684075B (en) Method for unloading computing tasks based on edge computing and cloud computing cooperation
CN111586696B (en) Resource allocation and unloading decision method based on multi-agent architecture reinforcement learning
CN110098969B (en) Fog computing task unloading method for Internet of things
Lee et al. An online secretary framework for fog network formation with minimal latency
CN111372314A (en) Task unloading method and task unloading device based on mobile edge computing scene
CN111475274B (en) Cloud collaborative multi-task scheduling method and device
WO2023040022A1 (en) Computing and network collaboration-based distributed computation offloading method in random network
CN111930436A (en) Random task queuing and unloading optimization method based on edge calculation
CN111182570A (en) User association and edge computing unloading method for improving utility of operator
CN114422349B (en) Cloud-edge-end-collaboration-based deep learning model training and reasoning architecture deployment method
EP2135430A2 (en) Hierarchically clustered p2p streaming system
EP3073732B1 (en) A method for allocating a video conferencing task to a processing device
CN103699447A (en) Cloud computing-based transcoding and distribution system for video conference
CN109947574A (en) A kind of vehicle big data calculating discharging method based on mist network
CN112650581A (en) Cloud-side cooperative task scheduling method for intelligent building
CN113064665A (en) Multi-server computing unloading method based on Lyapunov optimization
CN107302801B (en) QoE-oriented double-layer matching game method in 5G mixed scene
CN113992678A (en) Calculation migration method for offshore MEC load balancing and resource allocation joint optimization
CN113891477A (en) Resource allocation method based on MEC calculation task unloading in Internet of vehicles
CN112040512A (en) Fair-based fog computing task unloading method and system
Yao et al. Energy-aware task allocation for mobile IoT by online reinforcement learning
CN109547345B (en) Software-defined airborne network system and content-driven routing method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant