CN112596910B - Cloud computing resource scheduling method in multi-user MEC system - Google Patents

Cloud computing resource scheduling method in multi-user MEC system Download PDF

Info

Publication number
CN112596910B
CN112596910B CN202011586658.7A CN202011586658A CN112596910B CN 112596910 B CN112596910 B CN 112596910B CN 202011586658 A CN202011586658 A CN 202011586658A CN 112596910 B CN112596910 B CN 112596910B
Authority
CN
China
Prior art keywords
mec server
task
user
subtask
mec
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.)
Active
Application number
CN202011586658.7A
Other languages
Chinese (zh)
Other versions
CN112596910A (en
Inventor
卢建刚
付佳佳
曾瑛
亓峰
郑鸿远
吴赞红
施展
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.)
Electric Power Dispatch Control Center of Guangdong Power Grid Co Ltd
Original Assignee
Electric Power Dispatch Control Center of Guangdong Power Grid 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 Electric Power Dispatch Control Center of Guangdong Power Grid Co Ltd filed Critical Electric Power Dispatch Control Center of Guangdong Power Grid Co Ltd
Priority to CN202011586658.7A priority Critical patent/CN112596910B/en
Publication of CN112596910A publication Critical patent/CN112596910A/en
Application granted granted Critical
Publication of CN112596910B publication Critical patent/CN112596910B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G06F9/5038—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]

Abstract

The application discloses a cloud computing resource scheduling method in a multi-user MEC system, wherein under the condition that a mobile user locally makes an unloading decision of minimizing cost according to expense, under the scene that user subtasks have relevance of sequential execution and the preference degree of the user to money and computation delay time are different, a server profit maximization model taking the server task execution order as a variable is established, and then the profit maximization model is solved based on an ant colony algorithm to obtain an optimal task execution order and an optimal task partition strategy of profit maximization. The method and the system have the advantages that the calculation time and cost of the user to the task are met, meanwhile, the maximization of the benefits of the server provider is achieved, and the relationship between the user cost and the MEC server benefits can be well balanced.

Description

Cloud computing resource scheduling method in multi-user MEC system
Technical Field
The application relates to the technical field of mobile edge computing, in particular to a cloud computing resource scheduling method in a multi-user MEC system.
Background
With the popularization of 5G networks, new applications derived from the 5G networks, such as smart grids, internet of vehicles, telemedicine, virtual reality, augmented reality, and the like, are rapidly developed. But these new applications and services are difficult to deploy due to limited computing power and battery capacity of mobile devices and internet of things devices.
The intelligent power grid technology for constructing the air-ground cooperative communication network based on 5G can be used for simultaneously accessing a satellite network and a ground network, the user gateway supports a Software Defined Network (SDN), and MEC functions are deployed at the user gateway through Network Function Virtualization (NFV), so that a part of applications with high computation intensity and large uplink bandwidth occupation are directly completed at the edge, and the network time delay and the pressure of the center are reduced. Meanwhile, content caching, computing and unloading based on MEC can improve content distribution efficiency and processing rate, and service experience of users is improved.
In the architecture of the space-ground collaborative MEC, a user generates a computing task, firstly, whether a local computing resource can meet the requirement of the user is judged, and if the local computing resource can not meet the requirement of the user, the user decides whether to offload the local computing resource to an edge computing node in a satellite network or an edge computing node of a ground network through a collaborative offload strategy. When the corresponding edge computing node receives the computing task, if the computing task is in a busy state, the computing task is sent to a data center for processing. Otherwise, it will decide whether to process the computing task in cooperation with the surrounding edge computing nodes according to its own computing power, which depends on the cooperative task scheduling policy of the MEC server, i.e. cloud computing resource scheduling.
In practical processes, a phenomenon that multiple users commonly occupy computing resources of an edge server usually occurs, and because the computing resources of a mobile edge server are limited, an unloading request of multiple users cannot be accepted at the same time, which is called a user partition problem.
In order to solve the problem of user partition and cloud computing resource scheduling, an algorithm such as a SearchAdjust algorithm is proposed, and the user obtains the maximum benefit through a computing partition technology. In the case where multiple users compete for edge server-side computing resources and each user will make offloading decisions locally that minimize their own costs (e.g., by computing delays and spending weights and reducing costs), the MEC server provider needs to achieve maximization of its revenue by employing a suitable cloud computing resource scheduling approach. Therefore, a cloud computing resource scheduling method capable of better balancing the relationship between the user cost and the MEC server profit is to be proposed.
Disclosure of Invention
The application provides a cloud computing resource scheduling method in a multi-user MEC system, which is used for solving the technical problem that the relation between user cost and MEC server income cannot be well balanced in the prior art.
In view of this, a first aspect of the present application provides a cloud computing resource scheduling method in a multi-user MEC system, the multi-user MEC system including at least two mobile users and one MEC server, each mobile user sending an offload request to the MEC server, the offload request including a plurality of sub-tasks of the two mobile users, the plurality of sub-tasks having an association that is sequentially executed, including the steps of:
s1: the mobile user making locally an offloading decision to minimize cost based on spending, the spending including calculating a delay time spending and a monetary spending;
s2: establishing a maximum benefit model of the MEC server taking task execution order as a variable based on the offloading decision made by the mobile user and the preference degree of the mobile user for money and calculation delay time acquired in advance;
s3: solving the maximum benefit model through an ant colony algorithm to obtain an optimal task execution sequence and an optimal task partition strategy with maximized benefit;
s4: the MEC server executes the offload request according to the optimal task partition policy and the optimal task execution order.
Preferably, the step S1 specifically includes:
s101: recording N mobile subscriber sets sending unloading requests to the same MEC server at the same time asx i,j E {0,1} is represented as an offload decision, where 0 represents the local computation of subtask (i, j) and 1 represents the computation of subtask (i, j) at the MEC server;
in the method, in the process of the invention,for local calculation of delay time, G i,j CPU cycles, f, required to calculate subtasks (i, j) i Computing power for local;
the local calculation monetary cost is:
in the method, in the process of the invention,to calculate the monetary cost locally, beta i For mobile user +.>Preference degree for calculating delay time, +.>Calculating a delay time for the local;
in the middle ofCalculating time for MEC server, G i,j CPU cycles, f, required for computing task (i, j) c Computing capacity for the MEC server;
the waiting time of the subtasks (i, j) in the cloud is as follows:
in the method, in the process of the invention,waiting time in cloud for subtask (i, j), for subtask (i, j)>For the completion time of subtask (i, j,/-)>Is the completion time of subtask (i, j-1);
the MEC server calculates the delay time as:
the MEC server calculates the monetary cost as:
in the method, in the process of the invention,calculating monetary cost, beta, for MEC server i For mobile user +.>Preference degree for calculating delay time, +.>Calculating delay time for MEC server, +.>Waiting time of subtask (i, j) in cloud end, 1-beta i G for preference of mobile user for money i,j CPU cycles, μ required for computing task (i, j) i Calculating the cost of the resource needed to be delivered for a single CPU cycle of the MEC server occupied by the mobile subscriber;
s102: the mobile user delays the local calculationThe local calculation of the monetary cost +.>The MEC server calculates the time +.>The waiting time of the subtasks (i, j) in the cloud is->MEC server calculates delay timeAnd said MEC server calculating money spending +.>Uploading the money cost to the MEC server, comparing the money cost calculated by the local MEC server with the money cost calculated by the MEC server, and making a corresponding unloading decision, wherein the unloading decision is expressed as:
at the same time, when x i,j When=1, the MEC server calculates the delay time to satisfyWherein d i,j Executing the maximum tolerant time for the MEC server, and executing the maximum tolerant time d by the MEC server i,j The calculation formula of (2) is as follows:
wherein beta is i For mobile usersPreference degree for calculating delay time, +.>For local calculation of delay time, 1-beta i G for preference of mobile user for money i,j CPU cycles, x, required for computing task (i, j) i,j Mu for offloading decisions i The cost of the resource delivery required is calculated for a single CPU cycle of the mobile subscriber to occupy the MEC server.
Preferably, the maximum benefit model of the MEC server in the step S2 is converted into a MEC server benefit maximization problem, which is:
in the formula, defining the number of tasks in a resource occupation list of the MEC server as k; s is the execution order of the subtasks in the cloud, s= (S (1), S (2),..s (K)), where S (K) = (i, j), K represents the kth position of the jth subtask of user i in the cloud execution order; s is(s) j+1 (k)＝(i,j+1)；β s(k) ＝β i ；μ s(k) ＝μ i ；f s(k) ＝f i ；G s(k) ＝G i,j ；x s(k) ＝x i,j ；z s(k),s(o) ＝z (i,j),(i',j') ，s (o) ＝(i',j')，z (i,j),(i',j') Is the conception of the seedExecution order between transactions, where z (i,j),(i',j') =1 means that subtask (i, j) is performed before task (i ', j'), z (i,j),(i',j') =0 means that subtask (i ', j') is performed before (i, j); stc is the MEC server idle starting time; INF is infinity.
Preferably, the step S3 specifically includes:
s301: when the resources of the MEC server are not occupied, solving the optimal sub-task partition of each task;
s302: calculating a resource occupation list Lcro of the MEC server;
s303: searching conflict tasks of the MEC server from the resource occupation list Lcro, thereby forming a conflict task set Lcon;
s304: searching out the optimal order S of the execution of the conflict tasks in the resource occupation list Lcro through an ant colony algorithm con Then the best order S con The tasks of which the MEC server calculates the money cost is smaller than the local MEC server, are put into the task set S in turn * con In (a) and (b);
s305: updating cloud task execution orderRejecting conflict tasks in the conflict task set Lcon from the resource occupation list Lcro, and initializing the conflict task set Lcon and the task set S * con Judging whether the rejected resource occupation list Lcro has residual tasks, if so, executing the steps S303-S305, and if not, obtaining the optimal task execution sequence S of the MEC server;
s306: the tasks in the task optimal execution order S of the MEC server are MEC server computing tasks, and the remaining tasks are local computing tasks, so that an optimal task partition strategy of each mobile user is obtained.
Preferably, the step S303 specifically includes:
s3031: initializing relevant parameters and pheromone concentration of the ant colony algorithm, and setting the number m of ants, wherein the number m of ants is equal to the number of conflict tasks in the conflict task set Lcon;
s3032: inputting the starting time of each task in the resource occupation list Lcro to the maximum benefit modelMEC server calculation time ∈>MEC server executing maximum tolerance time +.>And +/income of each conflict task to the cloud end>
S3033: putting m ants at different departure points (i 1 ,j 1 ) Each ant selects the next arriving task with probability (i 2 ,j 2 ) Wherein, (i) 1 ,j 1 ),(i 2 ,j 2 )∈L con After execution of the task (i 1 ,j 1 ) Post execution task (i) 2 ,j 2 ) If (if)The added benefit of the MEC server is +.>Otherwise the MEC server increases the gain +.>
S3034: updating the pheromone concentration on the path after all ants pass through a round of path selection;
s3035: judging whether the preset maximum task cycle number is reached, if not, returning to the step S3033 to continue task cycleThe method comprises the steps of carrying out a first treatment on the surface of the If so, outputting the optimal order S of execution of the conflict tasks after finishing the task cycle of the ant colony algorithm con And its benefits.
Preferably, each ant in the step S3033 selects the task (i) to be arrived next with probability 2 ,j 2 ) The calculation formula of the probability of (a) is as follows:
wherein t is the number of task cycles,is path (i) 1 j 1 ,i 2 j 2 ) The pheromone concentration on the sample; />As heuristic function, alpha is information heuristic factor, beta is desired heuristic factor, alloweC k Task set C to be accessed, denoted as kth ant k The method comprises the steps of carrying out a first treatment on the surface of the Wherein heuristic function->The update rule of (2) is:
where revenue is the collection of fees to be delivered for each subtask.
Preferably, the update formula of the pheromone concentration in the step S3034 is:
in the method, in the process of the invention,is path (i) 1 j 1 ,i 2 j 2 ) The next time onTask circulation pheromone concentration to be updated, wherein ρ is the pheromone volatilization coefficient, and +.>Is path (i) 1 j 1 ,i 2 j 2 ) The pheromone concentration of the current task cycle,for all ants in task (i 1 ,j 1 ) And task (i) 2 ,j 2 ) An increased concentration of pheromone by releasing the pheromone on the connection path between them, wherein->In (1) the->For the kth ant in task (i 1 ,j 1 ) Subtasks (i) 2 ,j 2 ) The connection path between the two is used for releasing the pheromone and increasing the pheromone concentration, and the updating rule is as follows:
wherein Q is the intensity of pheromone.
From the above technical solutions, the embodiments of the present application have the following advantages:
the invention provides a cloud computing resource scheduling method in a multi-user MEC system, which comprises the steps of establishing a server profit maximization model taking a server task execution order as a variable under the scene that a user subtask has a sequential execution relevance and the user preference degree of money and computation delay time is different under the condition that a mobile user locally makes an unloading decision of minimizing cost according to cost, and then solving the profit maximization model based on an ant colony algorithm to obtain an optimal task execution order and an optimal task partition strategy of profit maximization. The method and the system have the advantages that the calculation time and cost of the user to the task are met, meanwhile, the maximization of the benefits of the server provider is achieved, and the relationship between the user cost and the MEC server benefits can be well balanced.
Drawings
Fig. 1 is a schematic structural diagram of a multi-user MEC system according to an embodiment of the present application;
fig. 2 is a flowchart of a cloud computing resource scheduling method in a multi-user MEC system according to an embodiment of the present application;
FIG. 3 is a graph of mean gain for MEC servers for two algorithms provided by embodiments of the present application;
fig. 4 is a graph of the calculated delay per user for two algorithms provided in embodiments of the present application.
Detailed Description
In order to make the present application solution better understood by those skilled in the art, the following description will clearly and completely describe the technical solution in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
As shown in fig. 1, the multi-user MEC system includes at least two mobile users and one MEC server, each mobile user sending an offload request to the MEC server, the offload request including a plurality of sub-tasks of the two mobile users, the plurality of sub-tasks having an association to be sequentially executed.
Currently, in situations where multiple mobile users are involved competing for MEC server computing resources and each user will make offloading decisions locally that minimize their own costs, it is difficult for the MEC server provider to achieve maximization of its revenue through appropriate resource scheduling policies.
Therefore, the invention provides a cloud computing resource scheduling method in a multi-user MEC system, which is based on the following three preconditions:
(1) Channel state information is known;
(2) The channel state remains unchanged during task offloading;
(3) Once it is decided to offload tasks to the MEC server, the mobile user will not stop offloading until offloading is complete.
For easy understanding, please refer to fig. 2, the present invention provides a cloud computing resource scheduling method in a multi-user MEC system, which includes the following steps:
s1: the mobile user makes an offloading decision to minimize cost locally based on costs including calculating delay time costs and monetary costs;
s2: based on the unloading decision made by the mobile user and the preference degree of the mobile user for money and calculation delay time, establishing a maximum benefit model of the MEC server taking the task execution order as a variable;
s3: solving a maximum benefit model through an ant colony algorithm to obtain an optimal task execution sequence and an optimal task partitioning strategy with maximized benefit;
s4: the MEC server performs offload requests according to the optimal task partitioning strategy and the optimal task execution order.
It should be noted that, in this embodiment, under the situation that the mobile user locally makes an offloading decision of minimizing cost according to the expense, and the user subtasks have relevance of sequential execution, and the preference degree of the user for money and calculation delay time is different, a server profit maximization model using the server task execution order as a variable is established, and then, the profit maximization model is solved based on the ant colony algorithm to obtain the best task execution order and best task partition strategy of profit maximization. The method and the system have the advantages that the calculation time and cost of the user on the task are met, and meanwhile the income of the server provider is maximized.
Further, the step S1 specifically includes:
s101: recording N mobile subscriber sets sending unloading requests to the same MEC server at the same time asx i,j E {0,1} is represented as an offload decision, where 0 represents the local computation of sub-task (i, j), and 1 represents the sub-task (i, j) is serviced at MECCalculating by a calculator;
let x ij When=0, the local calculation delay time is:
in the method, in the process of the invention,for local calculation of delay time, G i,j CPU cycles, f, required to calculate subtasks (i, j) i Computing power for local;
the local calculation monetary cost is:
in the method, in the process of the invention,to calculate the monetary cost locally, beta i For mobile user +.>Preference degree for calculating delay time, +.>Calculating a delay time for the local;
let x i，j When=1, the MEC server calculation time is:
in the middle ofCalculating time for MEC server, G i,j CPU cycles, f, required for computing task (i, j) c Computing capacity for the MEC server;
the waiting time of the subtasks (i, j) in the cloud is as follows:
in the method, in the process of the invention,waiting time in cloud for subtask (i, j), for subtask (i, j)>For the completion time of subtask (i, j,/-)>Is the completion time of subtask (i, j-1);
the MEC server calculates the delay time as:
the MEC server calculates the monetary cost as:
in the method, in the process of the invention,calculating monetary cost, beta, for MEC server i For mobile user +.>Preference degree for calculating delay time, +.>Calculating delay time for MEC server, +.>Waiting time of subtask (i, j) in cloud end, 1-beta i G for preference of mobile user for money i,j CPU cycles, μ required for computing task (i, j) i Calculating the cost of the resource needed to be delivered for a single CPU cycle of the MEC server occupied by the mobile subscriber;
s102: the mobile user will calculate the delay time locallyLocal calculation of money costs->MEC server calculation time ∈>Latency of subtasks (i, j) in cloud +.>MEC server calculates delay time +.>And MEC server calculates money costs ++>Uploading to the MEC server, comparing the local money cost calculation with the money cost calculation by the MEC server, and making a corresponding unloading decision, wherein the unloading decision is expressed as:
at the same time, when x i,j When=1, the MEC server calculates the delay time to satisfyWherein d i,j Executing the maximum tolerant time for the MEC server, and executing the maximum tolerant time d by the MEC server i,j The calculation formula of (2) is as follows:
wherein beta is i For mobile usersPreference degree for calculating delay time, +.>For local calculation of delay time, 1-beta i G for preference of mobile user for money i,j CPU cycles, x, required for computing task (i, j) i,j Mu for offloading decisions i The cost of the resource delivery required is calculated for a single CPU cycle of the mobile subscriber to occupy the MEC server.
It can be understood that the MEC server compares the cost of local computing and server computing to determine whether to offload, meanwhile, because the MEC server has limited computing resources and users have requirements on computing delay time, multiple users offload tasks to the MEC server to execute, and occupy server computing resources at the same time, so that the MEC server cannot meet the computing resource requirements of all offload users, therefore, the computing delay time of the MEC server needs to be less than the maximum tolerance time of the MEC server to execute, and the computing cost of the offload user server is ensured to be less than the local computing cost.
Further, the maximum benefit model of the MEC server in step S2 is converted into a MEC server benefit maximization problem, which is:
in the formula, defining the number of tasks in a resource occupation list of the MEC server as k; s is the execution order of the subtasks in the cloud, s= (S (1), S (2),..s (K)), where S (K) = (i, j), K represents the kth position of the jth subtask of user i in the cloud execution order; s is(s) j+1 (k)＝(i,j+1)；β s(k) ＝β i ；μ s(k) ＝μ i ；f s(k) ＝f i ；G s(k) ＝G i,j ；x s(k) ＝x i,j ；z s(k),s(o) ＝z (i,j),(i',j') ，s (o) ＝(i',j')，z (i,j),(i',j') Is the execution order among the subtasks, wherein z (i,j),(i',j') =1 means that subtask (i, j) is performed before task (i ', j'), z (i,j),(i',j') =0 means that subtask (i ', j') is performed before (i, j); stc is the MEC server idle starting time; INF is infinity.
It should be noted that, in this embodiment, the MEC server profit maximization problem is modeled as an optimization problem of the MEC server taking the task execution order as a variable, and is denoted as P1, and the constraint condition C1 ensures the execution order among the sub-tasks of a single mobile user; constraint C2 ensures MEC server execution order of subtasks of two different mobile users; constraint C3 ensures that the computational expense of the offloading task at the MEC server is less than the computational expense at the local site; constraint C4 limits unload variable x s(k) The value of 0 or 1 is adopted, so that the local calculation time of the subtasks, the waiting time of the MEC server and the calculation time of the server side are ensured to be more than 0; constraint C5 ensures that the computational expense of offloading to the MEC server is less than the local computational expense when the MEC server is idle.
Further, the step S3 specifically includes:
s301: when the resources of the MEC server are not occupied, solving the optimal sub-task partition of each task;
s302: calculating a resource occupation list Lcro of the MEC server;
s303: searching conflict tasks of the MEC server from the resource occupation list Lcro, thereby forming a conflict task set Lcon;
s304: searching out the optimal order S of the execution of the conflict tasks in the resource occupation list Lcro through an ant colony algorithm con Then the best order S con The tasks of which the MEC server calculates the money cost is smaller than the local MEC server, are put into the task set S in turn * con In (a) and (b);
s305: updating cloud task execution orderEliminating conflict tasks in the conflict task set Lcon from the resource occupation list Lcro, and initializing the conflict task set Lcon and the task set S * con Judging whether the rejected resource occupation list Lcro has residual tasks, if so, executing steps S303-S305, and if not, obtaining the optimal task execution sequence S of the MEC server;
s306: the tasks in the task optimal execution sequence S of the MEC server are MEC server computing tasks, and the remaining tasks are local computing tasks, so that an optimal task partition strategy of each mobile user is obtained.
It should be noted that, a plurality of tasks that use the computing resources of the MEC server at a time are called as conflicting tasks, and due to different preference degrees of money and computing delay time of users, the benefits of the MEC server can be improved while satisfying the demands of the users for unloading by adjusting the execution sequence of the conflicting tasks.
In this embodiment, to solve the optimization problem P1, in the worst case, the K ≡ that is needed at the K subtasks! In the permutation set, an execution order that makes the MEC server obtain the maximum benefit is found, so the P1 problem is a combinatorial optimization problem. Enumeration, approximation and heuristic algorithms are typically used to solve this type of problem. However, when the problem is large in scale, the enumeration method is too long in solving time, and an approximation algorithm is difficult to find out an accurate solution, so that the embodiment adopts an ant colony algorithm with stronger robustness and stronger global optimizing capability in a heuristic algorithm, and searches a task execution order which enables MEC service providers to obtain the maximum benefit from a plurality of arrangement sets.
Further, step S303 specifically includes:
s3031: initializing relevant parameters and pheromone concentration of an ant colony algorithm, and setting the number m of ants, wherein the number m of ants is equal to the number of conflict tasks in a conflict task set Lcon;
s3032: inputting the starting time of each task in the resource occupation list Lcro to the maximum benefit modelMEC server calculation time ∈>MEC server executing maximum tolerance time +.>Income of each conflict task to the cloud
S3033: putting m ants at different departure points (i 1 ,j 1 ) Each ant selects the next arriving task with probability (i 2 ,j 2 ) Wherein, (i) 1 ,j 1 ),(i 2 ,j 2 )∈L con After execution of the task (i 1 ,j 1 ) Post execution task (i) 2 ,j 2 ) If (if)The added benefit of the MEC server is +.>Otherwise the MEC server increases the gain +.>
S3034: updating the pheromone concentration on the path after all ants pass through a round of path selection;
s3035: judging whether the preset maximum task cycle times are reached, if not, returning to the step S3033 to continue task cycle; if so, outputting the optimal order S of execution of the conflict tasks after finishing the task cycle of the ant colony algorithm con And its benefits.
Further, each ant selects the next arriving task (i) with probability in step S3033 2 ,j 2 ) The calculation formula of the probability of (a) is as follows:
wherein t is the number of task cycles,is path (i) 1 j 1 ,i 2 j 2 ) The pheromone concentration on the sample; />As heuristic function, alpha is information heuristic factor, beta is desired heuristic factor, alloweC k Task set C to be accessed, denoted as kth ant k The method comprises the steps of carrying out a first treatment on the surface of the Wherein heuristic function->The update rule of (2) is:
where revenue is the collection of fees to be delivered for each subtask.
In this embodiment, α reflects the role of pheromone on ant pathways, and the information heuristic factor α=1, β reflects the relative importance of the heuristic function in guiding ant colony searches, with the heuristic factor β=5 being expected.
Further, the update formula of the pheromone concentration in step S3034 is:
in the method, in the process of the invention,is path (i) 1 j 1 ,i 2 j 2 ) The concentration of the pheromone to be updated in the next task cycle, ρ is the volatile coefficient of the pheromone,/->Is path (i) 1 j 1 ,i 2 j 2 ) The pheromone concentration of the current task cycle,for all ants in task (i 1 ,j 1 ) And task (i) 2 ,j 2 ) An increased concentration of pheromone by releasing the pheromone on the connection path between them, wherein->In (1) the->For the kth ant in task (i 1 ,j 1 ) Subtasks (i) 2 ,j 2 ) The connection path between the two is used for releasing the pheromone and increasing the pheromone concentration, and the updating rule is as follows:
wherein Q is the intensity of pheromone.
In the present embodiment, the pheromone volatilization coefficient ρ=0.1, and the pheromone intensity q=100.
In the prior art, only the computing partition technology is considered, so that the user obtains the maximum benefit, and the benefits of deployment and maintenance cost of MEC server recovery equipment are not considered.
The algorithm provided by the invention is simulated, the performance of the algorithm is verified, and the algorithm is compared with the SearchAdjust algorithm to perform comparison of average gain of the MEC server and average calculation delay of a user.
Simulation setting that each user has 5 subtasks to be offloaded, and the subtasks have relevance of sequential execution; the server side has two charging criteria for the user to choose from, each user randomly choosing one from the two charging criteria. The specific parameters are set as follows: the total number of users sending requests to the server is [5, 90]Within the range; two charging standards are u 0 ＝1×10 -9 Meta/cycles, u 1 ＝0.5×10 -9 Meta/cycles; CPU period required by each subtask is randomly valued at 0.1 GHz-1 GHz; the computing power of each device is randomly valued at 1-2 GHz. The SearchAdjus algorithm is then modified so that its optimized objective function is MEC server yield, and the modified SearchAdjus algorithm and the proposed algorithm are simulated and compared under the same parameters. During simulation, the number of ant colony is set to be the number of conflict tasks, the maximum number of iterations is 500, the information heuristic factor is 1, the expected heuristic factor is 5, the pheromone volatilization coefficient is 0.1, the pheromone intensity is 100, and MATLAB is used for simulation.
It can be seen from the graph of the gain curve of the MEC server of the two algorithms shown in fig. 3 and the graph of the average calculated delay per user shown in fig. 4:
(1) The gain of both algorithms increases with the increase of users, but the gain of the algorithm provided by the invention increases faster than the reference algorithm. When the number of users is 90 and fc=4 GHz, the gain of the algorithm provided by the invention is improved by 33.6% compared with the reference algorithm, and when the number of users is 90 and fc=12 GHz, the gain of the algorithm provided by the invention is improved by 49.9% compared with the reference algorithm. This is because with the increase in computing power, the algorithm of the present invention can process more tasks within a tolerable execution period of each task than the SearchAdjust algorithm.
(2) The average delay per user in both algorithms increases with the increase of users, when the number of users is 90 and fc=4 GHz, the average delay per user of the algorithm provided by the invention increases by 1.6% compared with the reference algorithm, and when the number of users is 90 and fc=12 GHz, the average delay per user of the algorithm provided by the invention increases by 6% compared with the reference algorithm.
Therefore, the algorithm provided by the invention greatly improves the benefit of the MEC server, realizes the maximization of the benefit of the MEC server, and can well balance the relationship between the user cost and the benefit of the MEC server.
The above embodiments are only for illustrating the technical solution of the present application, and are not limiting thereof; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.

Claims (5)

1. A cloud computing resource scheduling method in a multi-user MEC system, the multi-user MEC system comprising at least two mobile users and a MEC server, each mobile user sending an offload request to the MEC server, the offload request comprising a plurality of sub-tasks of the two mobile users, the plurality of sub-tasks having an association for sequential execution, comprising the steps of:
s1: the mobile user making locally an offloading decision to minimize cost based on spending, the spending including calculating a delay time spending and a monetary spending;
the S1 specifically comprises the following steps:
s101: recording N mobile subscriber sets sending unloading requests to the same MEC server at the same time asx i,j E {0,1} is represented as an offload decision, where 0 represents the local computation of subtask (i, j) and 1 represents the computation of subtask (i, j) at the MEC server;
let x i,j When=0, the local calculation delay time is:
in the method, in the process of the invention,for local calculation of delay time, G i,j CPU cycles, f, required to calculate subtasks (i, j) i Computing power for local;
the local calculation monetary cost is:
in the method, in the process of the invention,to calculate the monetary cost locally, beta i Preference degree for calculating delay time for mobile user i, < +.>Calculating delay time for the local, wherein j represents the serial number of the subtask;
let x i,j When=1, the MEC server calculation time is:
in the method, in the process of the invention,calculating time for MEC server, G i,j CPU cycles, f, required to calculate subtasks (i, j) c Computing capacity for the MEC server;
the waiting time of the subtasks (i, j) in the cloud is as follows:
in the method, in the process of the invention,waiting time in cloud for subtask (i, j), for subtask (i, j)>For the completion time of subtask (i, j,/-)>Is the completion time of subtask (i, j-1);
the MEC server calculates the delay time as:
the MEC server calculates the monetary cost as:
in the method, in the process of the invention,calculating monetary cost, beta, for MEC server i For the preference degree of the mobile user i for calculating the delay time,g for preference of mobile user for money i,j CPU cycles, μ required for computation of subtasks (i, j) i Calculating the cost of the resource needed to be delivered for a single CPU cycle of the MEC server occupied by the mobile subscriber;
s102: the mobile user delays the local calculationThe local calculation of the monetary cost +.>The MEC server calculates the time +.>The waiting time of the subtasks (i, j) in the cloud is->MEC server calculates delay time +.>And said MEC server calculating money spending +.>Uploading the money cost to the MEC server, comparing the money cost calculated by the local MEC server with the money cost calculated by the MEC server, and making a corresponding unloading decision, wherein the unloading decision is expressed as:
at the same time, when x i,j When=1, the MEC server calculates the delay time to satisfyWherein d i,j Executing the maximum tolerant time for the MEC server, and executing the maximum tolerant time d by the MEC server i,j The calculation formula of (2) is as follows:
wherein beta is i For the preference degree of the mobile user i for calculating the delay time,for local calculation of delay time, 1-beta i G for preference of mobile user for money i,j CPU cycles, x, required to calculate subtasks (i, j) i,j Mu for offloading decisions i Calculating the cost of the resource needed to be delivered for a single CPU cycle of the MEC server occupied by the mobile subscriber;
s2: establishing a maximum benefit model of the MEC server taking task execution order as a variable based on the offloading decision made by the mobile user and the preference degree of the mobile user for money and calculation delay time acquired in advance;
converting the maximum benefit model of the MEC server in the S2 into a MEC server benefit maximization problem, the MEC server benefit maximization problem being:
P1：
in the formula, defining the number of tasks in a resource occupation list of the MEC server as K; s is the execution order of the subtasks in the cloud, s= (S (1), S (2),..s (K)), where S (K) = (i, j), K represents the kth position of the jth subtask of user i in the cloud execution order; s is(s) j+1 (k)＝(i,j+1)；β s(k) ＝β i ；μ s(k) ＝μ i ；f s(k) ＝f i ；G s(k) ＝G i,j ；x s(k) ＝x i,j ；z s(k),s(o) ＝z (i,j),(i',j') ，s (o) ＝(i',j')，z (i,j),(i',j') Is the execution order among the subtasks, wherein z (i,j),(i',j') =1 means that subtask (i, j) is performed before subtask (i ', j'), z (i,j),(i',j') =0 means that subtask (i ', j') is performed before subtask (i, j); stc is the MEC server idle starting time; INF is infinity; p1 represents an optimization problem of the MEC server, wherein the MEC server profit maximization problem is modeled as an MEC server taking a task execution sequence as a variable, and C1, C2, C3, C4 and C5 all represent constraint conditions;
s3: solving the maximum benefit model through an ant colony algorithm to obtain an optimal task execution sequence and an optimal task partition strategy with maximized benefit;
s4: the MEC server executes the offload request according to the optimal task partition policy and the optimal task execution order.
2. The cloud computing resource scheduling method in a multi-user MEC system according to claim 1, wherein S3 specifically includes:
s301: when the resources of the MEC server are not occupied, solving the optimal sub-task partition of each task;
s302: calculating a resource occupation list Lcro of the MEC server;
s303: searching conflict tasks of the MEC server from the resource occupation list Lcro, thereby forming a conflict task set Lcon;
s304: searching out the optimal order S of the execution of the conflict tasks in the resource occupation list Lcro through an ant colony algorithm con Then the best order S con The tasks of which the MEC server calculates the money cost is smaller than the local MEC server, are put into the task set S in turn * con In (a) and (b);
s305: updating cloud task execution orderRejecting conflict tasks in the conflict task set Lcon from the resource occupation list Lcro, and initializing the conflict task set Lcon and the task set S * con Judging whether the rejected resource occupation list Lcro has residual tasks, if so, executing the S303-S305, and if not, obtaining the optimal execution sequence S of the tasks of the MEC server;
s306: the tasks in the task optimal execution order S of the MEC server are MEC server computing tasks, and the remaining tasks are local computing tasks, so that an optimal task partition strategy of each mobile user is obtained.
3. The cloud computing resource scheduling method in a multi-user MEC system according to claim 2, wherein S303 specifically includes:
s3031: initializing relevant parameters and pheromone concentration of the ant colony algorithm, and setting the number m of ants, wherein the number m of ants is equal to the number of conflict tasks in the conflict task set Lcon;
s3032: inputting the starting time of each task in the resource occupation list Lcro to the maximum benefit modelMEC server calculation time ∈>MEC server executing maximum tolerance time +.>And +/income of each conflict task to the cloud end>
S3033: putting m ants at different departure points (i 1 ,j 1 ) Each ant selects with probability the next arriving subtask (i 2 ,j 2 ) Wherein, (i) 1 ,j 1 ),(i 2 ,j 2 )∈L con Executing the subtask (i 1 ,j 1 ) Post-execution subtasks (i) 2 ,j 2 ) If (if)The added benefit of the MEC server is +.>Otherwise the MEC server increases the gain +.>
S3034: updating the pheromone concentration on the path after all ants pass through a round of path selection;
s3035: judging whether the preset maximum task cycle times are reached, if not, returning to the S3033 to continue task cycle; if so, outputting the optimal order S of execution of the conflict tasks after finishing the task cycle of the ant colony algorithm con And its benefits.
4. A method for scheduling cloud computing resources in a multi-user MEC system according to claim 3 wherein each ant in S3033 selects with probability the next arriving task (i 2 ,j 2 ) The calculation formula of the probability of (a) is as follows:
wherein t is the number of task cycles,is path (i) 1 j 1 ,i 2 j 2 ) The pheromone concentration on the sample; />As heuristic function, alpha is information heuristic factor, beta is desired heuristic factor, alloweC k Task set C to be accessed, denoted as kth ant k The method comprises the steps of carrying out a first treatment on the surface of the Wherein heuristic function->The update rule of (2) is:
where revenue is the collection of fees to be delivered for each subtask.
5. The cloud computing resource scheduling method in a multi-user MEC system according to claim 3, wherein the update formula of the pheromone concentration in S3034 is:
in the method, in the process of the invention,is path (i) 1 j 1 ,i 2 j 2 ) The concentration of the pheromone to be updated in the next task cycle, ρ is the volatile coefficient of the pheromone,/->Is path (i) 1 j 1 ,i 2 j 2 ) The pheromone concentration of the current task cycle, < >>For all ants in subtask (i 1 ,j 1 ) And subtask (i) 2 ,j 2 ) An increased concentration of pheromone by releasing the pheromone on the connection path between them, wherein->In (1) the->For the kth ant in subtask (i 1 ,j 1 ) And subtask (i) 2 ,j 2 ) The connection path between the two is used for releasing the pheromone and increasing the pheromone concentration, and the updating rule is as follows:
wherein Q is the intensity of pheromone.
CN202011586658.7A 2020-12-28 2020-12-28 Cloud computing resource scheduling method in multi-user MEC system Active CN112596910B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011586658.7A CN112596910B (en) 2020-12-28 2020-12-28 Cloud computing resource scheduling method in multi-user MEC system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011586658.7A CN112596910B (en) 2020-12-28 2020-12-28 Cloud computing resource scheduling method in multi-user MEC system

Publications (2)

Publication Number Publication Date
CN112596910A CN112596910A (en) 2021-04-02
CN112596910B true CN112596910B (en) 2024-02-20

Family

ID=75203753

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011586658.7A Active CN112596910B (en) 2020-12-28 2020-12-28 Cloud computing resource scheduling method in multi-user MEC system

Country Status (1)

Country Link
CN (1) CN112596910B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113568731B (en) * 2021-09-24 2021-12-28 苏州浪潮智能科技有限公司 Task scheduling method, chip and electronic equipment
CN114337783B (en) * 2021-12-30 2023-11-17 中国电子科技集团公司电子科学研究院 Space distributed edge computing device and business processing method

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109062682A (en) * 2018-06-29 2018-12-21 广东工业大学 A kind of resource regulating method and system of cloud computing platform
CN109388492A (en) * 2018-10-09 2019-02-26 浙江工业大学 A kind of mobile block chain optimization calculation force distribution method under multiple edge calculations server scenes based on simulated annealing
CN109981753A (en) * 2019-03-07 2019-07-05 中南大学 A kind of system and resource allocation methods of the edge calculations of the software definition of internet of things oriented
CN110099384A (en) * 2019-04-25 2019-08-06 南京邮电大学 Resource regulating method is unloaded based on side-end collaboration more MEC tasks of multi-user
CN110377353A (en) * 2019-05-21 2019-10-25 湖南大学 Calculating task uninstalling system and method
CN110647403A (en) * 2019-10-31 2020-01-03 桂林电子科技大学 Cloud computing resource allocation method in multi-user MEC system
CN111163519A (en) * 2019-12-27 2020-05-15 东北大学秦皇岛分校 Wireless body area network resource allocation and task unloading algorithm with maximized system benefits
CN111611080A (en) * 2020-05-22 2020-09-01 中国科学院自动化研究所 Edge computing task cooperative scheduling method, system and device

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109062682A (en) * 2018-06-29 2018-12-21 广东工业大学 A kind of resource regulating method and system of cloud computing platform
CN109388492A (en) * 2018-10-09 2019-02-26 浙江工业大学 A kind of mobile block chain optimization calculation force distribution method under multiple edge calculations server scenes based on simulated annealing
CN109981753A (en) * 2019-03-07 2019-07-05 中南大学 A kind of system and resource allocation methods of the edge calculations of the software definition of internet of things oriented
CN110099384A (en) * 2019-04-25 2019-08-06 南京邮电大学 Resource regulating method is unloaded based on side-end collaboration more MEC tasks of multi-user
CN110377353A (en) * 2019-05-21 2019-10-25 湖南大学 Calculating task uninstalling system and method
CN110647403A (en) * 2019-10-31 2020-01-03 桂林电子科技大学 Cloud computing resource allocation method in multi-user MEC system
CN111163519A (en) * 2019-12-27 2020-05-15 东北大学秦皇岛分校 Wireless body area network resource allocation and task unloading algorithm with maximized system benefits
CN111611080A (en) * 2020-05-22 2020-09-01 中国科学院自动化研究所 Edge computing task cooperative scheduling method, system and device

Also Published As

Publication number Publication date
CN112596910A (en) 2021-04-02

Similar Documents

Publication Publication Date Title
CN107911478B (en) Multi-user calculation unloading method and device based on chemical reaction optimization algorithm
CN109561148B (en) Distributed task scheduling method based on directed acyclic graph in edge computing network
CN110096362B (en) Multitask unloading method based on edge server cooperation
CN111813506B (en) Resource perception calculation migration method, device and medium based on particle swarm optimization
CN110647403B (en) Cloud computing resource allocation method in multi-user MEC system
CN110489176B (en) Multi-access edge computing task unloading method based on boxing problem
CN112596910B (en) Cloud computing resource scheduling method in multi-user MEC system
CN110740473B (en) Management method for mobile edge calculation and edge server
CN111711962B (en) Cooperative scheduling method for subtasks of mobile edge computing system
CN112286677A (en) Resource-constrained edge cloud-oriented Internet of things application optimization deployment method
CN113867843B (en) Mobile edge computing task unloading method based on deep reinforcement learning
CN112799823B (en) Online dispatching and scheduling method and system for edge computing tasks
CN109005211B (en) Micro-cloud deployment and user task scheduling method in wireless metropolitan area network environment
CN110048966B (en) Coflow scheduling method for minimizing system overhead based on deadline
CN113139639B (en) MOMBI-oriented smart city application multi-target computing migration method and device
Lorido-Botran et al. ImpalaE: Towards an optimal policy for efficient resource management at the edge
Li et al. Profit maximization for service placement and request assignment in edge computing via deep reinforcement learning
CN114168328A (en) Mobile edge node calculation task scheduling method and system based on federal learning
CN112130927B (en) Reliability-enhanced mobile edge computing task unloading method
Xu et al. Online learning algorithms for offloading augmented reality requests with uncertain demands in MECs
CN113741999A (en) Dependency-oriented task unloading method and device based on mobile edge calculation
CN117579701A (en) Mobile edge network computing and unloading method and system
CN116302578B (en) QoS (quality of service) constraint stream application delay ensuring method and system
CN109462879B (en) Admission control method and system
CN115190121B (en) Cross-region-based micro-service overload scheduling system, method and equipment

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