CN111694636A - Electric power Internet of things container migration method oriented to edge network load balancing - Google Patents

Electric power Internet of things container migration method oriented to edge network load balancing Download PDF

Info

Publication number
CN111694636A
CN111694636A CN202010392704.3A CN202010392704A CN111694636A CN 111694636 A CN111694636 A CN 111694636A CN 202010392704 A CN202010392704 A CN 202010392704A CN 111694636 A CN111694636 A CN 111694636A
Authority
CN
China
Prior art keywords
container
migration
edge
edge node
containers
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
CN202010392704.3A
Other languages
Chinese (zh)
Other versions
CN111694636B (en
Inventor
周俊
陈冰冰
刘强
许洪华
邵苏杰
王徐延
辛辰
夏伟栋
马子童
吴冠儒
丁达成
孔玥
魏玲燕
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.)
Beijing University of Posts and Telecommunications
State Grid Jiangsu Electric Power Co Ltd
Nanjing Power Supply Co of State Grid Jiangsu Electric Power Co Ltd
Original Assignee
Beijing University of Posts and Telecommunications
State Grid Jiangsu Electric Power Co Ltd
Nanjing Power Supply Co of State Grid Jiangsu Electric Power Co 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 Beijing University of Posts and Telecommunications, State Grid Jiangsu Electric Power Co Ltd, Nanjing Power Supply Co of State Grid Jiangsu Electric Power Co Ltd filed Critical Beijing University of Posts and Telecommunications
Priority to CN202010392704.3A priority Critical patent/CN111694636B/en
Publication of CN111694636A publication Critical patent/CN111694636A/en
Application granted granted Critical
Publication of CN111694636B publication Critical patent/CN111694636B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • 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
    • G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

The application discloses an electric power Internet of things container migration method for edge network load balancing, which comprises the steps of establishing a container information vector and a container deployment matrix of an edge network; carrying out load balancing detection on the edge network to obtain an overload edge node list; considering resource utilization balance, residual resource balance, network transmission delay and container migration halt time, and establishing a container migration model of load balance joint migration cost; and based on an improved ant colony system algorithm, carrying out container migration according to the container deployment matrix, the overload edge node list and the container migration model. The method and the device can effectively solve the problems that the business busyness degree difference between the edge nodes is obvious and the like caused by the time-space distribution of the business requests with limited edge node resources and obvious imbalance of the edge network.

Description

Electric power Internet of things container migration method oriented to edge network load balancing
Technical Field
The invention belongs to the technical field of edge network load balancing, and relates to a power internet of things container migration method for edge network load balancing.
Background
With the construction and development of smart power grids and the increasing popularization of interconnected equipment and sensors in smart cities, under the gradually-formed novel power internet of things service requirements of large connection, wide coverage, real-time and intellectualization, the cloud center computing and storage load is greatly improved due to massive transmission processing data, the network transmission pressure is increased, and the service processing delay is difficult to meet. The edge computing provides services for relieving computing and storage pressure of a cloud center and guaranteeing time delay requirements of power internet of things services by deploying edge nodes with communication access, service processing, data storage and other capabilities such as edge intelligent processing equipment or servers on the edge side closer to the user terminal. The terminal equipment communicates with the edge node in a wired mode, a WiFi mode or a 4G/5G mode and the like, and sends the task request to the edge node instead of a cloud platform, so that the service intelligent capacity of the terminal side is increased while the network data transmission quantity is reduced.
However, under the edge computing architecture, the power internet of things still faces a series of problems. The service busy degree difference among the edge nodes is obvious due to limited edge node resources and obviously unbalanced service request space-time distribution of the edge network, the connection and processing requirements of the service cannot be met in time due to overlarge load of part of the edge nodes, and the processing capacity of the rest edge nodes is idle and the resources are not fully utilized. Therefore, it is necessary to implement load balancing of the edge network so as to optimize resource utilization of the edge node. The virtual machine real-time migration is realized by migrating a running virtual machine from one physical machine to another physical machine, and simultaneously keeping the connection of a client or an application program, so that an effective technology for balancing loads among edge nodes and improving the resource utilization rate of the edge nodes is realized. However, the virtual machine has a slow start-up speed, and it is difficult to satisfy the QoS (Quality of Service) requirement of the delay-sensitive Service in the edge computing environment. Meanwhile, as the number of deployed virtual machines in the edge network increases, the performance of the virtual machines will significantly decrease.
To overcome the drawbacks of virtual machine migration, a lightweight virtualization technique called container has been widely used. Compared with a virtual machine, the container adopts layered storage with a mirror image layer and a data layer separated, and the container realizes isolation by controlling groups Cgroups (control groups) instead of a hypervisor, thereby supporting quick start, deployment and release. In addition, containers are lightweight, migration-enabled, as they are operating system level virtualization technologies that share the host operating system kernel. Therefore, a container migration mechanism facing edge network load balancing is provided, the load balancing of the edge network is realized on the premise of meeting the requirements of various services of the power internet of things, and the very important practical significance for minimizing the influence of the container migration on users is achieved.
In order to understand the development status of the existing container migration technology, the existing papers and patents are searched, compared and analyzed, and the following technical information with high relevance to the invention is screened out:
the technical scheme 1: a patent of "a load dynamic migration method in a container and virtual machine mixed cloud environment" with publication number CN110347498A, which relates to a dynamic migration method for load balancing of a container and virtual machine mixed cloud, and is mainly completed by three steps: firstly, selecting an overloaded server by using a static threshold value method; secondly, judging the reason of overload on the overload server: if the influence of the load of the virtual machine is large, selecting the virtual machine with the high load on the overload server, and adding the virtual machine into a list of virtual machines to be migrated; otherwise, predicting the CPU utilization rate of the container by a least square method, selecting the container with the container load in the rising situation, and adding the container into the list of the containers to be migrated; thirdly, if the overload is caused because the load of the virtual machine is high, migrating the virtual machine in the virtual machine list to be migrated to the target server by adopting a hybrid copy method; otherwise, judging whether the virtual machine exists on the target server: if not, a virtual machine is created, the container in the container list to be migrated is migrated to the newly created virtual machine, and if yes, the container in the container list to be migrated is directly migrated to the virtual machine of the target server.
In summary, technical scheme 1 provides a dynamic migration method for load balancing of a container and virtual machine hybrid cloud, where the method includes: selecting an overloaded server by using a static threshold value method; judging the reason of overload on an overload server, and selecting a container to be migrated and a virtual machine; and migrating the migration target to the target server based on a mixed copy method. According to the technical scheme 1, the virtual machines or container resources with larger overload influence are migrated by utilizing cosine correlation, so that the server cluster can better achieve load balance. However, neglecting the selection of the migration target server easily results in the phenomenon of load imbalance reappearing after migration.
The technical scheme 2 is as follows: a container dynamic migration method and system based on minimum migration volume, with publication number CN107992353A, relates to a container dynamic migration method for minimizing migration volume in a server cluster, which is mainly completed by five steps: firstly, judging whether a server cluster meets load balance or not based on a static threshold method, and if not, adding a server with a load exceeding a CPU utilization rate threshold into a hotspot server list; secondly, obtaining the growth characteristic of each container based on minimum multiplication on the basis of the acquired container memory historical data, and constructing a first growth characteristic priority list; thirdly, constructing a second priority list based on the current memory occupancy rate of the container, and adding the two priority lists according to a certain weight to obtain a third priority list; fourthly, if the memory occupancy of the current container is larger than a certain set threshold value, the container is directly taken as a container to be migrated and added into a migration list according to a minimum migration volume principle; otherwise, selecting the containers arranged according to the priority, and adding the container with the highest priority into the migration list; fifthly, migrating the container in the migration list to the target server according to a pre-copying method.
Technical scheme 2 provides a container dynamic migration method and system based on minimum migration volume, firstly judging whether server cluster loads are balanced, then carrying out priority ordering on containers to be migrated based on minimum triple multiplication and a minimum migration volume principle, and finally migrating the containers in a migration list to a target server according to a pre-copying method. According to the technical scheme 2, the containers are migrated and selected by fitting the growth rate of the containers on the source server, so that the cluster load balance is ensured, and the resource utilization rate is improved. However, the minimum migration amount is sought, and the migration times and thus the migration time are increased greatly.
Technical scheme 3: a patent of a container dynamic migration method based on energy consumption optimization with publication number CN110308973A relates to a container dynamic migration method based on energy consumption optimization, which is mainly completed by four steps: firstly, acquiring resource utilization rate information of each node; secondly, judging the container migration time based on the CPU, the memory, the bandwidth and the threshold value of the utilization rate of the I/O combined resource; thirdly, transferring and sequencing the containers to be transferred according to the CPU utilization rate, the memory utilization rate and the energy consumption; fourthly, performing ascending order arrangement according to the CPU, the memory and the energy consumption, and selecting the target edge node.
Technical scheme 3 provides a container dynamic migration method based on energy consumption optimization, and the method comprises the following steps: acquiring resource utilization rate information of each node, judging container migration time, performing migration sequencing on containers to be migrated, performing ascending sequencing on a CPU, a memory and energy consumption, and selecting a target edge node. The technical scheme 3 fully considers the migration opportunity, the selection of the container to be migrated and the selection of the target edge node. But the consideration factor is single, and the optimization target only considers one type of resources, so that the load balance of the whole cluster cannot be ensured.
Disclosure of Invention
In order to overcome the defects in the prior art, the application provides an electric power internet of things container migration method for edge network load balancing.
In order to achieve the above objective, the following technical solutions are adopted in the present application:
an electric power Internet of things container migration method facing edge network load balancing is characterized in that:
the method comprises the following steps:
step 1: establishing a container information vector and a container deployment matrix of an edge network;
step 2: carrying out load balancing detection on the edge network to obtain an overload edge node list;
and step 3: considering resource utilization balance, residual resource balance, network transmission delay and container migration halt time, and establishing a container migration model of load balance joint migration cost;
and 4, step 4: and based on an improved ant colony system algorithm, carrying out container migration according to the container deployment matrix, the overload edge node list and the container migration model.
The invention further comprises the following preferred embodiments:
preferably, the set of edge nodes in the edge network region in step 1 is
Figure BDA0002486429580000041
The collection of containers is
Figure BDA0002486429580000042
Each edge node comprises multiple types of resources, and the set of the resources in the power internet of things is
Figure BDA0002486429580000043
I. J, K are the number of containers, edge nodes and resource types in the electric power internet of things area respectively;
for any edge node
Figure BDA0002486429580000044
Its total amount of resources vector is
Figure BDA0002486429580000045
The resource utilization vector is
Figure BDA0002486429580000046
Wherein the content of the first and second substances,
Figure BDA0002486429580000047
represents njMaximum r that can be providedkThe amount of type resources is such that,
Figure BDA0002486429580000048
represents njUpper rkThe resource utilization rate of (2);
the edge nodes are connected in a wired or wireless mode, njAnd nj'Bandwidth in between, i.e. maximum transmission rate is Bandj,j'。
Preferably, in step 1, for any container
Figure BDA0002486429580000049
Its container information vector is
Figure BDA00024864295800000410
Wherein the content of the first and second substances,
Figure BDA00024864295800000411
denotes ciThe vector of resource requirements of (a),
Figure BDA00024864295800000412
denotes ciThe time at which the operation is started is,
Figure BDA00024864295800000413
representing the calculated time delay, Tol, under conditions that satisfy resource requirementsiIs a tolerable delay threshold;
the resource requirement of the container is the number of the virtual resource units after standardized processing;
the container deployment matrix in step 1 is a two-dimensional container deployment decision matrix X ═ X mapped from the container to the edge nodei,j]I×J；
Wherein the decision variable xi,jIndicating container ciWhether it has already been deployed at the edge node njIf deployed, x i,j1 is ═ 1; otherwise xi,j＝0；
Edge node njThe upper deployed container set is V (n)j)。
Preferably, step 2 comprises the steps of:
step 2.1: calculating the utilization rate of each type of resource of all edge nodes;
step 2.2: calculating the load of each edge node;
step 2.3: calculating the load difference between the edge nodes;
step 2.4: constructing a load difference matrix model among edge nodes and determining container migration triggering conditions;
step 2.5: judging whether container migration is triggered, if so, performing step 2.6, otherwise, returning to step 2.2;
step 2.6: and determining the overload edge nodes based on the triggering conditions in the step 2.4, and putting the overload edge nodes into an overload edge node list to obtain an overload edge node list.
Preferably, in step 2.1, njUpper rkResource utilization ratio of
Figure BDA0002486429580000051
The calculation formula is as follows:
Figure BDA0002486429580000052
preferably, in step 2.2, the edge node njThe load calculation formula of (a) is:
Figure BDA0002486429580000053
wherein, ηkIs rkIs on njThe weight of the load, and satisfies:
Figure BDA0002486429580000054
preferably, in step 2.3, any two edge nodes njAnd nj'The calculation formula of the load difference is as follows:
Figure BDA0002486429580000055
preferably, the load differentiation matrix module between edge nodes constructed in step 2.4 is HJ×J：
Figure BDA0002486429580000056
The container migration triggering conditions are as follows:
Figure BDA0002486429580000057
where σ is the upper bound of the load difference between edge nodes, ThrkIs the upper limit of the utilization rate of each type of resource.
Preferably, the container migration model established in step 3 is:
Figure BDA0002486429580000061
wherein:
Figure BDA0002486429580000062
Figure BDA0002486429580000063
Figure BDA0002486429580000064
wherein: theta and gamma are respectively the weight of the load balance degree of the edge network and the container migration cost;
Figure BDA0002486429580000065
for migrating decision variables, xi,j'For containers c after migrationiWhether to deploy at edge node njThe above step (1); if xi,j1 and xi,j'When 1, then there are
Figure BDA0002486429580000066
Indicating container ciFrom edge node njMove to nj'Otherwise
Figure BDA0002486429580000067
ΛiIs ciThe amount of transfer data generated during the migration process,
Figure BDA0002486429580000068
to give ciAllocated memory resources, Bj,j'Is njAnd nj'Bandwidth in between, i.e., maximum transmission rate;
Figure BDA0002486429580000069
representing post-migration edge node njUpper rkThe utilization rate of the resources of (a),
Figure BDA00024864295800000610
is represented by rkThe average resource utilization across all edge nodes,
Figure BDA00024864295800000611
indicates n after migration is completedjUpper rkThe amount of remaining available resources of the network,
Figure BDA00024864295800000612
is njUpper rkThe total amount of resources.
Preferably, step 4 comprises the steps of:
step 4.1: inputting a container deployment matrix and an overload edge node list;
step 4.2: calculating and sequencing the container migration priority on the overload edge node;
step 4.3: defining a container to be migrated as a container with the highest priority on the current overload edge node, and considering that the container to be migrated is not deployed on the overload edge node any more;
step 4.4: adding a container to be migrated into a container list to be migrated;
step 4.5: recalculating the load of the overload edge node after the container to be migrated is released, judging whether the load of the edge network is unbalanced or not based on the container migration triggering condition, if so, returning to the step 4.3, otherwise, performing the step 4.6;
step 4.6: selecting a target edge node for a container to be migrated according to the container migration model;
step 4.7: and finishing the migration scheduling of the container to be migrated based on the improved ant colony system algorithm, and outputting a migration result.
Preferably, in step 4.2, njUpper ciThe migration priority calculation formula is as follows:
Figure BDA0002486429580000071
wherein: λ, μ, ψ are weight factors based on the migration time, the number of times of migration, and the amount of migration data, respectively, and satisfy: λ + μ + ψ ═ 1;
Figure BDA0002486429580000072
is an edge node njUpper container ciBased on the migration probability of the migration time, the calculation formula is as follows:
Figure BDA0002486429580000073
wherein, ω iscpuAnd ωmemIs the weight of the impact of the pre-specified container CPU and memory utilization on migration downtime, and meets omegacpu+ωmem＝1；
Figure BDA0002486429580000074
Is an edge node njUpper container ciBased on the migration probability of the migration times, the calculation formula is as follows:
Figure BDA0002486429580000075
Figure BDA0002486429580000076
for the euclidean distance between the container load and the edge node load, the calculation formula is:
Figure BDA0002486429580000077
wherein the content of the first and second substances,
Figure BDA0002486429580000081
are respectively edge nodes njAnd a container ciUpper rkThe resource utilization rate of (2);
Figure BDA0002486429580000088
is rkTo njThe weight of load influence, and satisfy:
Figure BDA0002486429580000082
Figure BDA0002486429580000083
is an edge node njUpper container ciBased on the migration probability of the migration data volume, the calculation formula is as follows:
Figure BDA0002486429580000084
wherein the content of the first and second substances,
Figure BDA0002486429580000085
are respectively containers ciThe executed time and the computation time delay to meet the business resource requirement.
Preferably, in step 4.6, the selection mode of the target edge node is as follows:
Figure BDA0002486429580000086
preferably, in step 4.7, the ant colony system completes the migration scheduling of the container by simulating the ant foraging process, and specifically includes:
(1) mixing Ant AntlRandomly arranged in a container c to be migratediThe above step (1);
(2)Antlaccording to pheromone taui,jAnd heuristic information ηi,jSelecting a mapping relation tuple < ci,njI.e. ciIs deployed to njThe above step (1); then c is mixediPut in AntlTabu of Tabu tablelPerforming the following steps;
(3)Antlreturn to the container set C to be migratedmigRepeating the step (2) to complete the next migration allocation of the next container to be migrated to obtain a migration scheme, wherein all ants finish the allocation of the containers to be migrated once and are regarded as one iteration, and the algorithm is terminated after the maximum iteration number is reached.
Preferably, in step (2), the initial pheromone τ is0The values of (A) are:
Figure BDA0002486429580000087
wherein, | CmigI is the number of the containers to be migrated;
in selecting a new mapping relation tuple < ci,njAfter that, the ant updates the pheromone level of the traversal mapping relation based on the following local pheromone updating rule:
τi,j＝(1-ρl)τi,j(19)
where ρ islIs a local pheromone evaporation coefficient and satisfies rhol∈[0,1]，ρlThe larger the value of < c >i,njThe less pheromones remain above.
Preferably, in step (2), the heuristic information ηi,jThe calculation formula of (2) is as follows:
Figure BDA0002486429580000091
Figure BDA0002486429580000092
to aim atiMigration to njHeuristic information of the resulting migration costs,
Figure BDA0002486429580000093
comprises the following steps:
Figure BDA0002486429580000094
Figure BDA0002486429580000095
for the target edge node n after container migrationjThe heuristic information of the impact of (c),
Figure BDA0002486429580000096
comprises the following steps:
Figure BDA0002486429580000097
wherein the content of the first and second substances,
Figure BDA0002486429580000098
is ciTo rkThe resource requirements of (1).
Preferably, in step (2), the ants select the mapping relation tuples to traverse according to the following pseudo-random proportion rule:
Figure BDA0002486429580000099
wherein q is0Is between the interval [0,1]Q is generated from [0,1 ]]A random number in between;
when q is less than or equal to q0When the ant directly selects n which maximizes the product of the index α of the pheromone and the index β of the heuristic informationjAs c isiThe target edge node of (1); otherwise, the target edge node is selected based on the roulette rule of equation (25):
Figure BDA00024864295800000910
wherein p isi,jRepresents ciSelecting njAs the probability of its target edge node, Θl(i) To satisfy Ant of constraint conditionslSet of valid edge nodes of (c), Θl(i) Comprises the following steps:
Figure BDA0002486429580000101
preferably, in step (3), after all ants complete the construction of the migration solution, the quality of all currently constructed migration solutions is evaluated according to the objective function F, and the best one of the migration solutions is selected to execute the following global pheromone update rule to retain the experience of the global optimal solution:
Figure BDA0002486429580000102
where ρ isgUpdate coefficients for global pheromones and satisfy rhog∈[0,1]Δ τ is the increment of the extra pheromone, X+Is a global optimal solution in one iteration.
The beneficial effect that this application reached:
the container migration problem facing the edge network load balancing is expressed as a multi-target combination optimization problem under QoS constraint, a container migration model of load balancing combined migration cost is established by considering resource utilization balance degree, residual resource balance degree, network transmission delay and container migration halt time, and the container migration model is used for balancing the edge network load and minimizing the influence caused by container migration.
Determining Migration priority of a Container from two angles of resource correlation and service correlation, and designing an improved ant colony System Migration algorithm (CMDM-MACS, Container Migration-Based determination-creating Modified AntColony System) Based on a Container Migration Decision scheme CMDM (Container Migration-Based determination-creating) for solving the problem of Container Migration towards edge network load balancing under the scene of power Internet of things. By introducing a pseudo-random proportion rule and simultaneously combining local pheromone evaporation with global pheromone updating, the convergence speed is ensured while the algorithm exploration capacity is enhanced.
The method can effectively solve the problems of obvious difference of business busyness degrees among edge nodes and the like caused by limited edge node resources and obviously unbalanced business request space-time distribution of an edge network.
Drawings
Fig. 1 is a flowchart of an electric power internet of things container migration method for edge network load balancing according to the present application;
FIG. 2 is a schematic diagram of an improved ant colony system algorithm of the present application;
FIG. 3 is a graph comparing loads of edge nodes before and after container migration in the embodiment of the present application;
FIG. 4 is a graph comparing the load levels of the edge network before and after container migration in the example of the present application;
FIG. 5 is a graph of container migration costs in an embodiment of the present application.
Detailed Description
The present application is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present application is not limited thereby.
As shown in fig. 1, an electric power internet of things container migration method facing edge network load balancing in the present application includes the following steps:
step 1: establishing a container information vector and a container deployment matrix of an edge network;
in the specific embodiment of the present application, the set of edge nodes in the edge network region is
Figure BDA0002486429580000111
The collection of containers is
Figure BDA0002486429580000112
Each edge node comprises various resources such as a CPU, an internal memory, storage and the like, and the resources in the power internet of things are collected into
Figure BDA0002486429580000113
I. J, K are the regional contents of electric power Internet of thingsThe number of devices, edge nodes, resource types;
for any edge node
Figure BDA0002486429580000114
Its total amount of resources vector is
Figure BDA0002486429580000115
The resource utilization vector is
Figure BDA0002486429580000116
Wherein the content of the first and second substances,
Figure BDA0002486429580000117
represents njMaximum r that can be providedkThe amount of type resources is such that,
Figure BDA0002486429580000118
represents njUpper rkThe resource utilization rate of (2);
the edge nodes are connected in a wired or wireless mode, njAnd nj'Bandwidth in between, i.e. maximum transmission rate is Bandj,j'。
In step 1, for any container
Figure BDA0002486429580000119
Its container information vector is
Figure BDA00024864295800001110
Wherein the content of the first and second substances,
Figure BDA00024864295800001111
denotes ciThe vector of resource requirements of (a),
Figure BDA00024864295800001112
denotes ciThe time at which the operation is started is,
Figure BDA00024864295800001113
representing the calculated time delay, Tol, under conditions that satisfy resource requirementsiIs a tolerable delay threshold;
the resource requirement of the container is the number of the virtual resource units after standardized processing;
the container deployment matrix in step 1 is a two-dimensional container deployment decision matrix X ═ X mapped from the container to the edge nodei,j]I×J；
Wherein the decision variable xi,jIndicating container ciWhether it has already been deployed at the edge node njIf deployed, xi,j1 is ═ 1; otherwise xi,j0. Defining an edge node njThe upper deployed container set is V (n)j)。
Step 2: the method for carrying out load balancing detection on the edge network to obtain an overload edge node list comprises the following steps:
step 2.1: calculating the utilization rate of each type of resource, n, of all edge nodes simultaneouslyjUpper rkResource utilization ratio of
Figure BDA0002486429580000121
The calculation formula is as follows:
Figure BDA0002486429580000122
step 2.2: calculating the load of each edge node, edge node njThe load calculation formula of (a) is:
Figure BDA0002486429580000123
wherein, ηkIs rkIs on njThe weight of the load, and satisfies:
Figure BDA0002486429580000124
step 2.3: calculating the load difference between edge nodes, any two edge nodes njAnd nj'The calculation formula of the load difference is as follows:
Figure BDA0002486429580000125
step 2.4: constructing a load difference matrix model among edge nodes and determining a container migration triggering condition, wherein the constructed load difference matrix model among the edge nodes is HJ×J：
Figure BDA0002486429580000126
The container migration triggering conditions are as follows:
Figure BDA0002486429580000127
where σ is the upper bound of the load difference between edge nodes, ThrkIs the upper limit of the utilization rate of each type of resource.
Step 2.5: judging whether container migration is triggered, if so, performing step 2.6, otherwise, returning to step 2.2 to recalculate the load because the load of the edge network in the edge network is periodically changed, and realizing monitoring of the state of the edge network;
step 2.6: and (4) determining the overload edge nodes based on the container migration triggering conditions in the step 2.4, and putting the overload edge nodes into an overload edge node list to obtain an overload edge node list.
And step 3: considering resource utilization balance, residual resource balance, network transmission delay and container migration halt time, and establishing a container migration model of load balance joint migration cost;
defining migration decision variables
Figure BDA0002486429580000131
xi,j'For containers c after migrationiWhether to deploy at edge node njThe above step (1); two-dimensional container deployment decision matrix for container-to-edge node mapping after migration
Figure BDA0002486429580000132
If xi,j1 and xi,j'When 1, then there are
Figure BDA0002486429580000133
Indicating container ciFrom edge node njMove to nj'Otherwise
Figure BDA0002486429580000134
The migration cost is mainly composed of two parts: the network delay caused by the transmission of intermediate result data and final result data between two edge nodes and the migration time, i.e. down time, of the container itself. So as to be used as a container ciBy edge nodes njMigration to nj'The resulting migration cost can be expressed as:
Figure BDA0002486429580000135
wherein, ΛiIs ciThe amount of transfer data generated during the migration process,
Figure BDA0002486429580000136
to give ciAllocated memory resources, Bj,j'Is njAnd nj'The bandwidth in between, i.e., the maximum transmission rate.
The total migration cost is:
Figure BDA0002486429580000137
after all the containers to be migrated finish the migration operation, the load balance degree of the edge network is measured from two aspects of the resource utilization balance degree of the same type of resources between the edge nodes and the residual resource balance degree of different types of resources on the edge nodes.
Figure BDA0002486429580000138
Wherein the content of the first and second substances,
Figure BDA0002486429580000139
representing post-migration edge node njUpper rkThe utilization rate of the resources of (a),
Figure BDA00024864295800001310
is represented by rkThe average resource utilization across all edge nodes,
Figure BDA00024864295800001311
indicates n after migration is completedjUpper rkThe amount of remaining available resources of the network,
Figure BDA00024864295800001312
is njUpper rkThe total amount of resources.
To ensure reliable container migration, some QoS constraints relating to resources and latency are essential.
Figure BDA0002486429580000141
In summary, in order to balance the load balancing degree of the edge network after migration and the migration cost generated by the container in the migration process, the present patent describes the container migration problem to the load balancing of the edge network under the power internet of things scenario as a multi-objective optimization problem under Qos constraints and provides a container migration model:
Figure BDA0002486429580000142
and theta and gamma are respectively the weight of the load balance degree of the edge network and the container migration cost.
And 4, step 4: based on an improved ant colony system algorithm, container migration is carried out according to a container deployment matrix, an overload edge node list and a container migration model, and the method comprises the following steps:
step 4.1: inputting a container deployment matrix and an overload edge node list;
step 4.2: calculating and sequencing the container migration priority on the overload edge node;
the CPU and memory size of the container is a direct factor affecting migration downtime and resource loss. The smaller the CPU and memory utilization, the shorter the container migration downtime, and thus the less impact on performance and traffic QoS. So edge node njUpper container ciMigration probability based on migration time
Figure BDA0002486429580000143
Can be expressed as:
Figure BDA0002486429580000144
wherein, ω iscpuAnd ωmemIs the weight of the impact of the pre-specified container CPU and memory utilization on migration downtime, and meets omegacpu+ωmem＝1。
However, minimizing the migration time does not improve the load status of the edge network very well. Conversely, the overall migration down time may increase due to frequent container migration. Therefore, the method calculates the Euclidean distance between the container load and the edge node load while considering the single container migration time
Figure BDA0002486429580000145
To reduce the number of migrations.
Figure BDA0002486429580000151
Wherein the content of the first and second substances,
Figure BDA0002486429580000152
are respectively edge nodes njAnd a container ciUpper rkThe resource utilization rate of (2);
Figure BDA0002486429580000153
is rkTo njThe weight of load influence, and satisfy:
Figure BDA0002486429580000154
so edge node njUpper container ciMigration probability based on migration times
Figure BDA0002486429580000155
Can be expressed as:
Figure BDA0002486429580000156
from the perspective of the traffic carried by the container, the container migration necessarily generates network overhead for data transmission between any two edge nodes based on the traffic execution result. The size of the transmitted data volume is mainly influenced by the execution progress of the service, i.e. the more the execution progress of the service is, the smaller the generated relative data volume is.
So edge node njUpper container ciMigration probability based on migration data volume
Figure BDA0002486429580000157
Can be expressed as:
Figure BDA0002486429580000158
wherein the content of the first and second substances,
Figure BDA0002486429580000159
are respectively containers ciThe executed time and the computation time delay to meet the business resource requirement.
In summary, the following steps: n isjUpper ciThe migration priority of (d) may be expressed as:
Figure BDA00024864295800001510
wherein λ, μ, ψ are weighting factors based on the migration time, the number of times of migration, and the amount of migration data, respectively, and satisfy: λ + μ + ψ ═ 1.
Step 4.3: defining a container to be migrated as a container with the highest priority on the current overload edge node, and considering that the container to be migrated is not deployed on the overload edge node any more;
step 4.4: adding a container to be migrated into a container list to be migrated;
step 4.5: recalculating the load of the overload edge node after the container to be migrated is released, judging whether the load of the edge network is unbalanced or not based on the container migration triggering condition, if so, returning to the step 4.3, otherwise, performing the step 4.6;
step 4.6: selecting a target edge node for a container to be migrated according to the container migration model;
the selection mode of the target edge node is as follows:
Figure BDA0002486429580000161
step 4.7: and finishing the migration scheduling of the container to be migrated based on the improved ant colony system algorithm, and outputting a migration result.
In order to obtain a global optimal container migration decision result, an ant colony optimization system algorithm aiming at a discrete problem is designed based on a CMDM scheme.
The ant colony system completes the migration scheduling of the container by simulating the foraging process of ants, as shown in fig. 2, specifically including:
(1) mixing Ant AntlRandomly arranged in a container c to be migratediThe above step (1);
(2)Antlaccording to pheromone taui,jAnd heuristic information ηi,jSelecting a mapping relation tuple < ci,njI.e. ciIs deployed to njThe above step (1); then c is mixediPut in AntlTabu of Tabu tablelPerforming the following steps;
initial pheromone tau0The values of (A) are:
Figure BDA0002486429580000162
wherein, | CmigIs | toThe number of migration containers;
for pheromone updating in an Ant Colony System (ACS) algorithm, the application provides a pheromone updating rule combining local and global.
In selecting a new mapping relation tuple < ci,njAfter that, the ant updates the pheromone level of the traversal mapping relation based on the following local pheromone updating rule:
τi,j＝(1-ρl)τi,j(19)
where ρ islIs a local pheromone evaporation coefficient and satisfies rhol∈[0,1]，ρlThe larger the value of < c >i,njThe less pheromones remain above.
Heuristic information η based on the CMDM scheme proposed by the present patenti,jMainly according to ciMigration to njMigration cost and n after migrationjIs calculated to obtain.
Heuristic information ηi,jThe calculation formula of (2) is as follows:
Figure BDA0002486429580000171
Figure BDA0002486429580000172
to aim atiMigration to njHeuristic information of the resulting migration costs,
Figure BDA0002486429580000173
comprises the following steps:
Figure BDA0002486429580000174
Figure BDA0002486429580000175
for the target edge node n after container migrationjThe heuristic information of the impact of (c),
Figure BDA0002486429580000176
intended to avoid njSimultaneous balancing of resource overload njOf different types of resources. Therefore, it is
Figure BDA0002486429580000177
Comprises the following steps:
Figure BDA0002486429580000178
wherein the content of the first and second substances,
Figure BDA0002486429580000179
is n after migration is completedjUpper rkThe remaining available resources of the network are,
Figure BDA00024864295800001710
is ciTo rkThe resource requirements of (a) are determined,
Figure BDA00024864295800001711
is njUpper rkThe total amount of resources of;
in step (2), ants tend to select tuples with the largest pheromones and heuristic information. However, in order to avoid trapping in local optima, ants select a mapping relation tuple to traverse according to the following pseudo-random proportion rule:
Figure BDA00024864295800001712
wherein q is0Is between the interval [0,1]Q is generated from [0,1 ]]A random number in between;
when q is less than or equal to q0When the ant directly selects n which maximizes the product of the index α of the pheromone and the index β of the heuristic informationjAs c isiThe target edge node of (1); otherwise, the target edge node is selected based on the roulette rule of equation (25):
Figure BDA00024864295800001713
wherein p isi,jRepresents ciSelecting njAs the probability of its target edge node, Θl(i) To satisfy Ant of constraint conditionslSet of valid edge nodes of (c), Θl(i) Comprises the following steps:
Figure BDA0002486429580000181
(3)Antlreturn to the container set C to be migratedmigRepeating the step (2) to complete the next migration allocation of the next container to be migrated to obtain a migration scheme, wherein all ants finish the allocation of the containers to be migrated once and are regarded as one iteration, and the algorithm is terminated after the maximum iteration number is reached.
In the step (3), after all ants finish constructing the migration scheme, the quality of all currently constructed migration schemes is evaluated according to the objective function F, and the best one of the migration schemes is selected to execute the following global pheromone updating rules so as to keep the experience of the global optimal solution:
Figure BDA0002486429580000182
where ρ isgUpdate coefficients for global pheromones and satisfy rhog∈[0,1]Δ τ is the increment of the extra pheromone, X+Is a global optimal solution in one iteration.
The examples of the invention are as follows:
the patent presents numerical results to verify the performance of the proposed solution. In the simulation, the heterogeneous edge network environment is a rectangular area of 10kmx5km, the number of edge devices is 10, the number of containers is 30-210, and the positions of the containers to be deployed are randomly generated in the area. The resource types of the edge nodes are considered as follows: computing, memory and storage resources. The number of CPU cores of the edge devices is 32, the value ranges of the capacities of the memory and the storage resources are respectively [16,32] and [100,300] GB, and the value range of the transmission rate between the edge devices is [100,300] Mb/s. The resource requirement of the container is randomly generated, the value ranges of the CPU resource requirement, the memory resource requirement and the storage resource requirement of the container are respectively [1,3] core, [1,4] and [5,15] GB, and the service delay constraint of the service carried by the container is randomly set from 50ms to 500 ms.
Fig. 3 shows the load change of each edge node before and after container migration. As can be seen from fig. 3, the load of the edge node EN7 is already more than 80% before performing container migration, while the load of EN1, EN2, EN6 is only less than 30%. Under the condition that the load of the edge network is seriously unbalanced, as the terminal service request under the scene of the power internet of things has certain regularity and predictability, the container virtualization layer of the edge node with a large load is congested and the queuing delay is greatly increased along with the passage of time, so that the QoS (quality of service) requirements of services in a part of containers cannot be met, and the capability of the edge node with a small load cannot be fully utilized. After the CMDM _ MACS is executed, the loads of the edge nodes in the edge network are relatively balanced, the execution pressure of individual busy edge nodes is relieved, the possibility that the service QoS requirement is not satisfied is reduced, and meanwhile, the resource utilization of idle edge nodes is optimized.
Fig. 4 shows a comparison of the degree of load balancing of the edge network before and after container migration. Wherein, the load balancing specifically comprises: the utilization balance of any resource by different edge nodes and the balance of the remaining resources of different types of resources in any edge node. Smaller values of load balancing represent more balanced load at each edge node in the edge network. As can be seen from fig. 4, as the number of containers is increased, the CMDM policy of the present application guarantees load balancing of the edge network while improving the system migration cost.
FIG. 5 illustrates container migration costs under a CMDM migration selection policy. In general, under the condition that the total amount of the edge node resources is fixed, along with the continuous increase of the number of containers, the resource occupancy rate of the edge node is increased, the load difference degree of the edge network is obvious, and further more containers need to be migrated to the relatively idle edge nodes, so that the delay constraint of services in the containers is ensured. Therefore, as the number of containers increases, the migration cost in the system tends to increase linearly. The QoS aware-based container selection strategy proposed in this patent tends to migrate containers that can significantly reduce the load while considering reducing the single migration time, so as the number of containers increases, the CMDM migration strategy exhibits better performance.
The present applicant has described and illustrated embodiments of the present invention in detail with reference to the accompanying drawings, but it should be understood by those skilled in the art that the above embodiments are merely preferred embodiments of the present invention, and the detailed description is only for the purpose of helping the reader to better understand the spirit of the present invention, and not for limiting the scope of the present invention, and on the contrary, any improvement or modification made based on the spirit of the present invention should fall within the scope of the present invention.

Claims (17)

1. An electric power Internet of things container migration method facing edge network load balancing is characterized in that:
the method comprises the following steps:
step 1: establishing a container information vector and a container deployment matrix of an edge network;
step 2: carrying out load balancing detection on the edge network to obtain an overload edge node list;
and step 3: considering resource utilization balance, residual resource balance, network transmission delay and container migration halt time, and establishing a container migration model of load balance joint migration cost;
and 4, step 4: and based on an improved ant colony system algorithm, carrying out container migration according to the container deployment matrix, the overload edge node list and the container migration model.
2. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 1, wherein the method comprises the following steps:
step 1 the set of edge nodes in the edge network region is
Figure FDA0002486429570000011
The collection of containers is
Figure FDA0002486429570000012
Each edge node comprises multiple types of resources, and the set of the resources in the power internet of things is
Figure FDA0002486429570000013
I. J, K are the number of containers, edge nodes and resource types in the electric power internet of things area respectively;
for any edge node
Figure FDA0002486429570000014
Its total amount of resources vector is
Figure FDA0002486429570000015
The resource utilization vector is
Figure FDA0002486429570000016
Wherein the content of the first and second substances,
Figure FDA0002486429570000017
represents njMaximum r that can be providedkThe amount of type resources is such that,
Figure FDA0002486429570000018
represents njUpper rkThe resource utilization rate of (2);
the edge nodes are connected in a wired or wireless mode, njAnd nj'Bandwidth in between, i.e. maximum transmission rate is Bandj,j'。
3. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 2, characterized in that:
in step 1, for any container
Figure FDA0002486429570000019
Its container information vector is
Figure FDA00024864295700000110
Wherein the content of the first and second substances,
Figure FDA00024864295700000111
denotes ciThe vector of resource requirements of (a),
Figure FDA00024864295700000112
denotes ciThe time at which the operation is started is,
Figure FDA00024864295700000113
representing the calculated time delay, Tol, under conditions that satisfy resource requirementsiIs a tolerable delay threshold;
the resource requirement of the container is the number of the virtual resource units after standardized processing;
the container deployment matrix in step 1 is a two-dimensional container deployment decision matrix X ═ X mapped from the container to the edge nodei,j]I×J；
Wherein the decision variable xi,jIndicating container ciWhether it has already been deployed at the edge node njIf deployed, xi,j1 is ═ 1; otherwise xi,j＝0；
Edge node njThe upper deployed container set is V (n)j)。
4. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 3, wherein the method comprises the following steps:
the step 2 comprises the following steps:
step 2.1: calculating the utilization rate of each type of resource of all edge nodes;
step 2.2: calculating the load of each edge node;
step 2.3: calculating the load difference between the edge nodes;
step 2.4: constructing a load difference matrix model among edge nodes and determining container migration triggering conditions;
step 2.5: judging whether container migration is triggered, if so, performing step 2.6, otherwise, returning to step 2.2;
step 2.6: and determining the overload edge nodes based on the triggering conditions in the step 2.4, and putting the overload edge nodes into an overload edge node list to obtain an overload edge node list.
5. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 4, wherein the method comprises the following steps:
in step 2.1, njUpper rkResource utilization ratio of
Figure FDA0002486429570000021
The calculation formula is as follows:
Figure FDA0002486429570000022
6. the method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 5, wherein the method comprises the following steps:
in step 2.2, edge node njThe load calculation formula of (a) is:
Figure FDA0002486429570000023
wherein, ηkIs rkIs on njThe weight of the load, and satisfies:
Figure FDA0002486429570000024
7. the method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 6, wherein the method comprises the following steps:
in step 2.3, any two edge nodes njAnd nj'The calculation formula of the load difference is as follows:
Figure FDA0002486429570000031
8. the method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 7, wherein the method comprises the following steps:
the load differentiation matrix module between edge nodes constructed in step 2.4 is HJ×J：
Figure FDA0002486429570000032
The container migration triggering conditions are as follows:
Figure FDA0002486429570000033
where σ is the upper bound of the load difference between edge nodes, ThrkIs the upper limit of the utilization rate of each type of resource.
9. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 3, wherein the method comprises the following steps:
the container migration model established in the step 3 is as follows:
Figure FDA0002486429570000034
wherein:
Figure FDA0002486429570000035
Figure FDA0002486429570000041
Figure FDA0002486429570000042
Figure FDA0002486429570000043
wherein: theta and gamma are respectively the weight of the load balance degree of the edge network and the container migration cost;
Figure FDA0002486429570000044
for migrating decision variables, xi,j'For containers c after migrationiWhether to deploy at edge node njThe above step (1); if xi,j1 and xi,j'When 1, then there are
Figure FDA0002486429570000045
Indicating container ciFrom edge node njMove to nj'Otherwise
Figure FDA0002486429570000046
ΛiIs ciThe amount of transfer data generated during the migration process,
Figure FDA0002486429570000047
to give ciAllocated memory resources, Bj,j'Is njAnd nj'Bandwidth in between, i.e., maximum transmission rate;
Figure FDA0002486429570000048
representing post-migration edge node njUpper rkThe utilization rate of the resources of (a),
Figure FDA0002486429570000049
is represented by rkThe average resource utilization across all edge nodes,
Figure FDA00024864295700000410
indicates n after migration is completedjUpper rkThe amount of remaining available resources of the network,
Figure FDA00024864295700000411
is njUpper rkThe total amount of resources.
10. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 9, wherein the method comprises the following steps:
step 4 comprises the following steps:
step 4.1: inputting a container deployment matrix and an overload edge node list;
step 4.2: calculating and sequencing the container migration priority on the overload edge node;
step 4.3: defining a container to be migrated as a container with the highest priority on the current overload edge node, and considering that the container to be migrated is not deployed on the overload edge node any more;
step 4.4: adding a container to be migrated into a container list to be migrated;
step 4.5: recalculating the load of the overload edge node after the container to be migrated is released, judging whether the load of the edge network is unbalanced or not based on the container migration triggering condition, if so, returning to the step 4.3, otherwise, performing the step 4.6;
step 4.6: selecting a target edge node for a container to be migrated according to the container migration model;
step 4.7: and finishing the migration scheduling of the container to be migrated based on the improved ant colony system algorithm, and outputting a migration result.
11. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 10, wherein the method comprises the following steps:
in step 4.2, njUpper ciThe migration priority calculation formula is as follows:
Figure FDA0002486429570000051
wherein: λ, μ, ψ are weight factors based on the migration time, the number of times of migration, and the amount of migration data, respectively, and satisfy: λ + μ + ψ ═ 1;
Figure FDA0002486429570000052
is an edge node njUpper container ciBased on the migration probability of the migration time, the calculation formula is as follows:
Figure FDA0002486429570000053
wherein, ω iscpuAnd ωmemIs the weight of the impact of the pre-specified container CPU and memory utilization on migration downtime, and meets omegacpu+ωmem＝1；
Figure FDA0002486429570000054
Is an edge node njUpper container ciBased on the migration probability of the migration times, the calculation formula is as follows:
Figure FDA0002486429570000055
Figure FDA0002486429570000056
for the euclidean distance between the container load and the edge node load, the calculation formula is:
Figure FDA0002486429570000057
wherein the content of the first and second substances,
Figure FDA0002486429570000058
are respectively edge nodes njAnd a container ciUpper rkThe resource utilization rate of (2);
Figure FDA0002486429570000059
is rkTo njThe weight of load influence, and satisfy:
Figure FDA0002486429570000061
Figure FDA0002486429570000062
is an edge node njUpper container ciBased on the migration probability of the migration data volume, the calculation formula is as follows:
Figure FDA0002486429570000063
wherein the content of the first and second substances,
Figure FDA0002486429570000064
are respectively containers ciThe executed time and the computation time delay to meet the business resource requirement.
12. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 10, wherein the method comprises the following steps:
in step 4.6, the selection mode of the target edge node is as follows:
Figure FDA0002486429570000065
13. the method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 10, wherein the method comprises the following steps:
in step 4.7, the ant colony system completes the migration scheduling of the container by simulating the ant foraging process, and specifically comprises the following steps:
(1) mixing Ant AntlRandomly arranged in a container c to be migratediThe above step (1);
(2)Antlaccording to pheromone taui,jAnd heuristic information ηi,jSelecting a mapping relation tuple < ci,njI.e. ciIs deployed to njThe above step (1); then c is mixediPut in AntlTabu of Tabu tablelPerforming the following steps;
(3)Antlreturn to the container to be migrated list CmigRepeating the step (2) to complete the next migration allocation of the next container to be migrated to obtain a migration scheme, wherein all ants finish the allocation of the containers to be migrated once and are regarded as one iteration, and the algorithm is terminated after the maximum iteration number is reached.
14. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 13, wherein the method comprises the following steps:
in step (2), the initial pheromone τ0The values of (A) are:
Figure FDA0002486429570000071
wherein, | CmigI is the number of the containers to be migrated;
in selecting a new mapping relation tuple < ci,njAfter that, the ant updates the pheromone level of the traversal mapping relation based on the following local pheromone updating rule:
τi,j＝(1-ρl)τi,j(19)
where ρ islIs a local pheromone evaporation coefficient and satisfies rhol∈[0,1]，ρlThe larger the value of < c >i,njThe less pheromones remain above.
15. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 13, wherein the method comprises the following steps:
in step (2), heuristic information ηi,jThe calculation formula of (2) is as follows:
Figure FDA0002486429570000072
Figure FDA0002486429570000073
to aim atiMigration to njHeuristic information of the resulting migration costs,
Figure FDA0002486429570000074
comprises the following steps:
Figure FDA0002486429570000075
Figure FDA0002486429570000076
for the target edge node n after container migrationjThe heuristic information of the impact of (c),
Figure FDA0002486429570000077
comprises the following steps:
Figure FDA0002486429570000078
wherein the content of the first and second substances,
Figure FDA0002486429570000079
is ciTo rkThe resource requirements of (1).
16. The method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 13, wherein the method comprises the following steps:
in the step (2), ants select mapping relation tuples to traverse according to the following pseudo-random proportion rules:
Figure FDA00024864295700000710
wherein q is0Is between the interval [0,1]Q is generated from [0,1 ]]A random number in between;
when q is less than or equal to q0When the ant directly selects n which maximizes the product of the index α of the pheromone and the index β of the heuristic informationjAs c isiThe target edge node of (1); otherwise, the target edge node is selected based on the roulette rule of equation (25):
Figure FDA0002486429570000081
wherein p isi,jRepresents ciSelecting njAs the probability of its target edge node, Θl(i) To satisfy Ant of constraint conditionslSet of valid edge nodes of (c), Θl(i) Comprises the following steps:
Figure FDA0002486429570000082
17. the method for migrating the containers of the internet of things of electric power oriented to edge network load balancing according to claim 13, wherein the method comprises the following steps:
in the step (3), after all ants finish constructing the migration scheme, the quality of all currently constructed migration schemes is evaluated according to the objective function F, and the best one of the migration schemes is selected to execute the following global pheromone updating rules so as to keep the experience of the global optimal solution:
Figure FDA0002486429570000083
where ρ isgUpdate coefficients for the global pheromone, andsatisfy rhog∈[0,1]Δ τ is the increment of the extra pheromone, X+Is a global optimal solution in one iteration.
CN202010392704.3A 2020-05-11 2020-05-11 Electric power Internet of things container migration method oriented to edge network load balancing Active CN111694636B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010392704.3A CN111694636B (en) 2020-05-11 2020-05-11 Electric power Internet of things container migration method oriented to edge network load balancing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010392704.3A CN111694636B (en) 2020-05-11 2020-05-11 Electric power Internet of things container migration method oriented to edge network load balancing

Publications (2)

Publication Number Publication Date
CN111694636A true CN111694636A (en) 2020-09-22
CN111694636B CN111694636B (en) 2022-08-23

Family

ID=72477138

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010392704.3A Active CN111694636B (en) 2020-05-11 2020-05-11 Electric power Internet of things container migration method oriented to edge network load balancing

Country Status (1)

Country Link
CN (1) CN111694636B (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112445573A (en) * 2020-11-04 2021-03-05 许继集团有限公司 Edge Internet of things agent resource scheduling method and device based on standby mechanism
CN113259469A (en) * 2021-06-02 2021-08-13 西安邮电大学 Edge server deployment method, system and storage medium in intelligent manufacturing
CN113760541A (en) * 2021-07-29 2021-12-07 国网河南省电力公司信息通信公司 Method and device for distributing edge resources
CN113847926A (en) * 2021-09-18 2021-12-28 上海电机学院 Real-time path planning method based on edge micro-service cooperation
CN114039980A (en) * 2021-11-08 2022-02-11 郑州轻工业大学 Low-delay container migration path selection method and system facing edge collaborative computing
CN114077498A (en) * 2021-11-20 2022-02-22 郑州轻工业大学 Method and system for selecting and transferring calculation load facing to mobile edge calculation
CN114095358A (en) * 2021-11-23 2022-02-25 南方电网深圳数字电网研究院有限公司 Dynamic network-oriented virtual network function deployment method and device
CN114599072A (en) * 2022-02-14 2022-06-07 东南大学 Time-varying demand-driven dense WLAN terminal association management method
CN114723236A (en) * 2022-03-17 2022-07-08 广州汉光电气股份有限公司 Power grid edge calculation boundary method
US11388106B2 (en) * 2020-10-30 2022-07-12 EMC IP Holding Company LLC Method, device, and computer program product for edge resource aggregation
CN115118729A (en) * 2022-06-28 2022-09-27 中国电信股份有限公司 Container migration method, system and storage medium
CN116991580A (en) * 2023-07-27 2023-11-03 上海沄熹科技有限公司 Distributed database system load balancing method and device
CN117032996A (en) * 2023-10-09 2023-11-10 湖南中青能科技有限公司 Power metadata management method and system

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019007420A1 (en) * 2017-07-07 2019-01-10 中兴通讯股份有限公司 Load balance scheduling method and device, and computer readable storage medium
CN109600269A (en) * 2019-01-21 2019-04-09 云南电网有限责任公司信息中心 A kind of cloud management platform based on DCOS
CN109714400A (en) * 2018-12-12 2019-05-03 华南理工大学 A kind of energy optimization resource scheduling system and its method towards container cluster
US20190253490A1 (en) * 2016-10-31 2019-08-15 Huawei Technologies Co., Ltd. Resource load balancing control method and cluster scheduler
CN110457131A (en) * 2019-07-31 2019-11-15 华中科技大学 Electric system supercomputer platform method for scheduling task based on Docker container

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190253490A1 (en) * 2016-10-31 2019-08-15 Huawei Technologies Co., Ltd. Resource load balancing control method and cluster scheduler
WO2019007420A1 (en) * 2017-07-07 2019-01-10 中兴通讯股份有限公司 Load balance scheduling method and device, and computer readable storage medium
CN109714400A (en) * 2018-12-12 2019-05-03 华南理工大学 A kind of energy optimization resource scheduling system and its method towards container cluster
CN109600269A (en) * 2019-01-21 2019-04-09 云南电网有限责任公司信息中心 A kind of cloud management platform based on DCOS
CN110457131A (en) * 2019-07-31 2019-11-15 华中科技大学 Electric system supercomputer platform method for scheduling task based on Docker container

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11388106B2 (en) * 2020-10-30 2022-07-12 EMC IP Holding Company LLC Method, device, and computer program product for edge resource aggregation
CN112445573A (en) * 2020-11-04 2021-03-05 许继集团有限公司 Edge Internet of things agent resource scheduling method and device based on standby mechanism
CN113259469A (en) * 2021-06-02 2021-08-13 西安邮电大学 Edge server deployment method, system and storage medium in intelligent manufacturing
CN113259469B (en) * 2021-06-02 2022-10-14 西安邮电大学 Edge server deployment method, system and storage medium in intelligent manufacturing
CN113760541A (en) * 2021-07-29 2021-12-07 国网河南省电力公司信息通信公司 Method and device for distributing edge resources
CN113760541B (en) * 2021-07-29 2023-06-30 国网河南省电力公司信息通信公司 Method and device for allocating edge resources
CN113847926A (en) * 2021-09-18 2021-12-28 上海电机学院 Real-time path planning method based on edge micro-service cooperation
CN113847926B (en) * 2021-09-18 2024-01-19 上海电机学院 Real-time path planning method based on edge microservice collaboration
CN114039980A (en) * 2021-11-08 2022-02-11 郑州轻工业大学 Low-delay container migration path selection method and system facing edge collaborative computing
CN114077498B (en) * 2021-11-20 2023-03-28 郑州轻工业大学 Method and system for selecting and transferring calculation load facing to mobile edge calculation
CN114077498A (en) * 2021-11-20 2022-02-22 郑州轻工业大学 Method and system for selecting and transferring calculation load facing to mobile edge calculation
CN114095358A (en) * 2021-11-23 2022-02-25 南方电网深圳数字电网研究院有限公司 Dynamic network-oriented virtual network function deployment method and device
CN114599072A (en) * 2022-02-14 2022-06-07 东南大学 Time-varying demand-driven dense WLAN terminal association management method
CN114599072B (en) * 2022-02-14 2024-02-06 东南大学 Time-varying demand-driven dense WLAN terminal association management method
CN114723236A (en) * 2022-03-17 2022-07-08 广州汉光电气股份有限公司 Power grid edge calculation boundary method
CN115118729A (en) * 2022-06-28 2022-09-27 中国电信股份有限公司 Container migration method, system and storage medium
CN116991580A (en) * 2023-07-27 2023-11-03 上海沄熹科技有限公司 Distributed database system load balancing method and device
CN117032996A (en) * 2023-10-09 2023-11-10 湖南中青能科技有限公司 Power metadata management method and system
CN117032996B (en) * 2023-10-09 2023-12-22 湖南中青能科技有限公司 Power metadata management method and system

Also Published As

Publication number Publication date
CN111694636B (en) 2022-08-23

Similar Documents

Publication Publication Date Title
CN111694636B (en) Electric power Internet of things container migration method oriented to edge network load balancing
CN105491138B (en) Distributed load scheduling method based on load rate graded triggering
Bezerra et al. A load balancing scheme for massively multiplayer online games
CN112039965A (en) Multitask unloading method and system in time-sensitive network
CN110445866A (en) Task immigration and collaborative load-balancing method in a kind of mobile edge calculations environment
Ma et al. Container migration mechanism for load balancing in edge network under power Internet of Things
CN107220108A (en) A kind of method and system for realizing cloud data center load balancing
Sedaghat et al. Decentralized cloud datacenter reconsolidation through emergent and topology-aware behavior
WO2020134133A1 (en) Resource allocation method, substation, and computer-readable storage medium
CN113342409A (en) Delay sensitive task unloading decision method and system for multi-access edge computing system
CN108089918B (en) Graph computation load balancing method for heterogeneous server structure
CN113382074A (en) Micro-service load balancing optimization method based on dynamic feedback
Gomathi et al. Multi-objective optimization of energy aware virtual machine placement in cloud data center
CN110597598B (en) Control method for virtual machine migration in cloud environment
CN115022926A (en) Multi-objective optimization container migration method based on resource balance
CN111176784A (en) Virtual machine integration method based on extreme learning machine and ant colony system
Hayat et al. Efficient energy utilization in cloud fog environment
Dziurzanski et al. Value-based allocation of docker containers
CN112954012B (en) Cloud task scheduling method based on improved simulated annealing algorithm of load
CN113360245A (en) Internet of things equipment task downloading method based on mobile cloud computing deep reinforcement learning
CN113329432A (en) Edge service arrangement method and system based on multi-objective optimization
CN113535397A (en) Resource allocation method for intelligent vehicle edge computing network
CN110308965B (en) Rule-based heuristic virtual machine distribution method and system for cloud data center
CN110430236A (en) A kind of method and dispatching device of deployment business
Naik et al. Developing a cloud computing data center virtual machine consolidation based on multi-objective hybrid fruit-fly cuckoo search algorithm

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