A task allocation method considering task delay and server cost in mobile edge computing networks

Abstract

The present invention relates to a task allocation method considering task delay and server cost in a mobile edge computing network, and the method comprises the following steps: S1, obtaining network configuration information of a system model; S2, generating a task set, obtaining the calculation, transmission delay and server execution cost of the task set; S3, randomly initializing and generating an initial particle swarm of the task set; S4, constructing a fitness evaluation function according to the total cost of the calculation delay, the transmission delay and the server execution cost; S5, using a multi-objective gravitational search algorithm, introducing a convergence factor to update particle speed and position; S6, using a crossover mutation method to replace the position information of particles; S7, looping and iterating steps S5-S6 to obtain the task allocation strategy with the minimum total cost. The method of the present invention proposes a multi-objective gravitational search algorithm method to minimize the user task execution delay and server calculation cost.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a task allocation method considering task delay and server cost in a mobile edge computing network. Background Art

[0002] The emergence of the Internet of Things has led to the widespread use of more and more mobile user devices. However, the limited battery life of mobile devices and the low latency requirements of these applications have increased the demand for new network models. Mobile edge computing deploys computing and storage resources at the edge of mobile networks, providing a service environment and cloud computing capabilities for mobile networks, thereby providing users with ultra-low latency, high-bandwidth network services.

[0003] Computation offloading is one of the key technologies. It means that the terminal device transfers part or all of its computing tasks to the cloud computing environment to solve the deficiencies of mobile devices in terms of resource storage, computing performance and energy efficiency. However, since mobile edge computing servers usually have limited computing resources, they cannot meet the computing needs of all users at the same time. Therefore, in the scenario of collaborative computing offloading between cloud servers and edge servers, formulating offloading decision-making plans has become a hot issue. Summary of the invention

[0004] Based on the above-mentioned shortcomings and deficiencies in the prior art, one of the objects of the present invention is to at least solve one or more of the above-mentioned problems in the prior art. In other words, one of the objects of the present invention is to provide a task offloading and resource allocation method for a mobile edge computing network that meets one or more of the above-mentioned requirements.

[0005] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:

[0006] A task allocation method considering task delay and server cost in a mobile edge computing network, the method comprising the steps of:

[0007] S1. Obtain network configuration information of a system model, where the system model consists of several base station groups, a mobile cloud server, and user local servers belonging to a certain base station group. Each base station group includes a base station, several user local servers, and an edge computing server;

[0008] S2. Generate a task set containing priority constraint relationships, obtain the computing delay and server execution cost of the task set in the edge computing server and the mobile cloud server according to the network configuration information, and obtain the transmission delay of each task in the task set between the edge computing server and the mobile cloud server;

[0009] S3, randomly initialize and generate an initial particle swarm of task sets, assign tasks in the task set to the user's local server, edge computing server or mobile cloud server, each task is a particle, and encode the particle position of each task according to its assigned server;

[0010] S4, constructing a fitness evaluation function according to the total cost of computing delay, transmission delay and server execution cost, and calculating the fitness function value of each particle according to the initial particle swarm and particle position coding;

[0011] S5. Use the multi-objective gravitational search algorithm, introduce a convergence factor when calculating the resultant force of the multi-objective gravitational search algorithm, and update the particle velocity and position;

[0012] S6. Use the crossover mutation method to set a crossover probability. Each time a particle updates its position, it generates a random number. If the random number generated by the particle is less than the crossover probability, the original position information of the particle is replaced with the position information of a particle in the current particle group. And compare the change in the gravity of the particle before and after the replacement. If the gravity becomes larger, the position information after the replacement is replaced;

[0013] S7. Set the number of loops, and iterate steps S5-S6 according to the number of loops to obtain a task allocation strategy with the minimum total overhead.

[0014] As a preferred solution, the computing delay of each task in the task set on the edge computing server and the mobile cloud server is calculated by dividing the computing amount of each task by the number of CPU cycles that the edge computing server and the mobile cloud server can provide per second.

[0015] As a preferred solution, if two tasks in a task set that have a successor relationship are on the same server, the transmission delay is 0.

[0016] As a preferred solution, in step S2, the task set determines the energy consumption of the server in combination with the hardware architecture of the server when calculating the energy consumption of the user's local server, edge computing server, and mobile cloud server.

[0017] As a preferred solution, the average calculation delay of the task processing on the server in step S3 is obtained by adding the upload delay of the task, the calculation delay of the task on the server and the download delay after the task calculation is completed.

[0018] As a preferred solution, the server execution cost is calculated based on the computing delay of the task on the server, the billing cycle of the server, the computing power of the server, the minimum server computing power and the basic price of the server.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The method of the present invention is aimed at task offloading in a mobile edge computing system, proposes a mobile edge computing task computing offloading model, and proposes a multi-objective gravitational search algorithm method. Through coordinated computing between a mobile edge computing server and a mobile cloud server, different edge servers can communicate through a resource pool formed by their base stations, thereby optimizing the gravitational search method and minimizing user task execution delays and server computing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of a method for offloading calculations in a mobile edge computing network taking into account task priorities according to an embodiment of the present invention;

[0022] Figure 2 is a schematic diagram of the structure of a system model of an embodiment of the present invention;

[0023] Figure 3 It is a schematic diagram of a task model diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to more clearly illustrate the embodiments of the present invention, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings and other implementation methods can be obtained based on these accompanying drawings without creative work.

[0025] Embodiment: This embodiment provides a method for offloading calculations in a mobile edge computing network taking into account task priorities, and its flow chart is as follows: Figure 1 As shown: First, perform step S1 to obtain the network configuration information of the system model. The structure of the system model is as follows Figure 2 As shown in the figure, it consists of several base station groups, a mobile cloud server and user local servers belonging to a certain base station group. Each base station group includes a base station, several user local servers and an edge computing server. A server can only process one task at a time. When there are too many user-side task sets, the base station group to which the user belongs can communicate with other base station groups in the system model for collaborative computing. The scheduling plan is centrally decided by the mobile cloud server.

[0026] The network configuration information of the system model consists of the information of each server, P = {p i |1≤i≤M} is the server set in the system model, where p 1 For the user's local server, p 2 The edge computing server of the base station group where the user is located, p Mis a mobile cloud server, and the rest are other mobile cloud servers in the same system model as the base station group to which the user belongs. When considering allocating all tasks to mobile cloud servers and edge computing servers, use f E , f C They represent the computing power of the edge computing server and the mobile cloud server, that is, the number of CPU cycles that can be provided per second. The network configuration information also includes transmission parameters such as the fixed transmission power of the user's local server, the channel gain between the user's local server and the base station, and the channel bandwidth.

[0027] After obtaining the network configuration information, proceed to step S2, generate a task set containing a priority constraint relationship, obtain the computing delay and server execution cost of the task set in the user's local server, edge computing server, and mobile cloud server according to the network configuration information, and obtain the transmission delay of each task in the task set between the user's local server, edge computing server, and mobile cloud server;

[0028] Specifically, the priority constraint relationship between task sets is used as follows Figure 3 The directed acyclic graph model representation of the task model graph, Figure 3 The left side of the figure shows the task model of a single user, and the right side shows the task model of multiple users. The task model is named G=<V,E> . Where V is the task set, V = {v i |1≤i≤N}. E is a set of directed edges between tasks, indicating the priority constraint relationship between tasks. ij ∈E means that in task v i After the task is completed, you can execute the task v j Each task v i ={d i ,g i}, where d i The CPU cycles required to complete the task, g i Enter the data size for the task.

[0029] For a known directed acyclic graph model, task v i The set of predecessor nodes is pre(v i ), the successor node set is suc(v i ), the layer value of each task in the task set can be determined by the following formula:

[0030]

[0031] h(i) means that if task v i If there is no predecessor node, it is used as the first node. If there is a predecessor node, it is used as the next node.

[0032] According to the network configuration information obtained in S1, the computing delay of the task on each server can be obtained. This embodiment provides a specific calculation method for the computing delay:

[0033] The computation delay of the i-th task on the edge computing server is:

[0034] The computational delay of the i-th task on the mobile cloud server is:

[0035] Combined, the computational delay of the i-th task on the server is: Where α and β are 0-1 variables, α=1 means that the task is offloaded to the mobile edge server for processing, and β=1 means that the task is offloaded to the mobile cloud server for processing.

[0036] The transmission delay of tasks between the user's local server, edge computing server, and mobile cloud server is calculated using the following method:

[0037] During the transmission process, if the previous node and the next node are assigned to the same server, the transmission delay is zero. Otherwise, j Go to the next task v i The transmission delay is:

[0038]

[0039] in is a 0-1 decision variable, Represents subtask v i Processed by the user side, Represents subtask v i It is processed by the edge computing server under the base station to which the user belongs. Represents subtask v i There are three cases where the arrows are directed to the direction of task offloading. 0 ∈{p 2 ,…p M-1}, r E is the transmission speed between base stations in the resource pool, r U It is the transmission speed from base station to mobile cloud server.

[0040] On the user side, the completion time of the calculation task is: Where S i The time when the task starts executing.

[0041] According to the exponential pricing model, task v i On the server l The execution cost on is:

[0042]

[0043] Where σ represents a random variable used to generate different pricing models, is the computation delay, τ is the server billing period, and f k For server p l The computing power, f min is the minimum server computing capacity in the instance scenario, V based Indicates the base price of the server.

[0044] From this, we can get the total cost calculation formula for the server:

[0045]

[0046] Among them B i,l represents a 0-1 uninstall decision, B i,l =1 means subtask v i Offload to server l , cost i,l Represented as task v i Offload to server l execution costs.

[0047] Combining the computational delay, transmission delay, and server execution cost of the above task set in the user's local server, edge computing server, and mobile cloud server, the model of the overall optimization problem is:

[0048]

[0049]

[0050] Where λ is the gravity factor, which is used to balance the two optimization objectives of latency and cost. It is expressed as a normalization factor, which keeps the two optimization objectives of total delay and total cost at the same order of magnitude.

[0051] Then, step S3 is performed to randomly initialize the initial particle swarm of the generated task set, and the tasks in the task set are respectively assigned to the user's local server, edge computing server or mobile cloud server, each task is a particle, and the particle position of each task is encoded according to the server to which it is assigned; the particle position This corresponds to a complete solution to the problem. In the initialization phase of the method, the particle positions are randomly generated in the interval (0,1), and the particle dimension is N. d , the number of particles is N p .

[0052] Then, step S4 is performed to construct a fitness evaluation function according to the total cost of the calculation delay, the transmission delay and the server execution cost, and the fitness function value of each particle is calculated according to the initial particle group and the particle position code;

[0053] For a given particle position, its fitness value can be expressed as:

[0054] The gravity calculation formula of particle i is:

[0055] Among them, best(t) represents the maximum fitness value in the current particle swarm, and worst(t) represents the minimum fitness value in the current particle swarm.

[0056] Then, step S5 is performed, the particle position is iterated using a multi-objective gravitational search algorithm, a convergence factor is introduced when calculating the resultant force of the multi-objective gravitational search algorithm, and the particle speed and position are updated.

[0057] In order to improve the accuracy of the search, first initialize the gravity constant G 0 , the gravitational factor G(t) in the tth iteration can be expressed as:

[0058]

[0059] Here, γ is a constant in the interval (0,1).

[0060] In the tth iteration, particle i is subject to the gravitational force F of particle j ij It can be expressed as:

[0061]

[0062] Among them, R ij (t) is the Euclidean distance, and δ is a very small constant.

[0063] In order to avoid local optimality, the convergence factor kbest is introduced in the multi-objective gravitational search algorithm. The force acting on particle i can be expressed as:

[0064]

[0065] Among them, rand 1 is a random number generated in the interval [0,1], and K is the kbest particles with the largest mass in the current particle swarm.

[0066] In the tth iteration, the acceleration of particle i is:

[0067]

[0068] The calculation formulas for the velocity and position of particle i are:

[0069]

[0070]

[0071] In order to update the new solution of the particle swarm and compare the fitness value, step S6 is also performed, using the crossover mutation method to set a crossover probability. Each time a particle updates its position, a random number is generated. If the random number generated by the particle is less than the crossover probability, the original position information of the particle is replaced with the position information of a particle in the current particle swarm. The change in the gravity of the particle before and after the replacement is compared. If the gravity becomes larger, the replaced position information is retained.

[0072] The specific implementation method is as follows:

[0073] C1: Calculate the optimal solution AL in the current particle swarm best ; Select particle A from the particle group in order i .

[0074] C2: Generated random number rand 3 ,(rand 3 ∈(0,1)), and compare it with the crossover probability p c If the particle is smaller than , go to step C3; if it is larger, skip the particle and go to step C1.

[0075] C3: Let U i =A i ; Generate a random integer N l ,1 <N l <N d -l',

[0076] C4: Compare fitness (U i ) and fitness(A i ), if greater than, then A i =U i ; If less than A i =A i

[0077] C5: Determine whether the crossover operation of particles in the particle group has been completed. If not, jump to step C1; if yes, end the loop.

[0078] Finally, step S7 is performed, an iteration number is set, and steps S4-S5 are iterated repeatedly until the iteration number is reached. Then, the optimal server for each task allocation is determined using the historical optimal position of the particle finally determined, and the task allocation strategy with the minimum total cost is obtained.

[0079] The present invention further improves the GSA method, with the optimization goal of minimizing user execution delay and server computing cost, and offloading multiple users with the same subtask set to different servers. Different edge computing servers can communicate through the resource pool formed by their base stations. In order to avoid falling into the local optimal solution, a crossover factor is introduced when calculating the resultant force, and a crossover operation is introduced after each particle position update.

[0080] It should be noted that the above embodiments are only detailed descriptions of the preferred embodiments and principles of the present invention. For ordinary technicians in this field, there will be changes in the specific implementation methods based on the ideas provided by the present invention, and these changes should also be regarded as the scope of protection of the present invention.

Claims

1. A task allocation method considering task delay and server cost in a mobile edge computing network, characterized in that: The method comprises the steps of: S1. Obtain network configuration information of a system model, wherein the system model consists of several base station groups, a mobile cloud server, and a user local server belonging to one of the base station groups, and each of the base station groups includes a base station, several user local servers, and an edge computing server; S2, generating a task set including a priority constraint relationship, obtaining the computing delay and server execution cost of the task set in the edge computing server and the mobile cloud server according to the network configuration information, and obtaining the transmission delay of each task in the task set between the edge computing server and the mobile cloud server; S3, randomly initializing and generating an initial particle swarm of the task set, respectively assigning the tasks in the task set to the user local server, the edge computing server or the mobile cloud server, each of the tasks being a particle, and performing particle position encoding on the tasks according to the servers to which they are assigned; S4, constructing a fitness evaluation function according to the total overhead of the calculation delay, the transmission delay and the server execution cost, and calculating the fitness function value of each particle according to the initial particle swarm and the particle position code; S5, using a multi-objective gravitational search algorithm, introducing a convergence factor when calculating the resultant force of the multi-objective gravitational search algorithm, and updating the particle velocity and position; S6. Use the crossover mutation method to set a crossover probability. Each time a particle updates its position, it generates a random number. If the random number generated by the particle is less than the crossover probability, the original position information of the particle is replaced with the position information of a particle in the current particle group, and the change of the gravity of the particle before and after the replacement is compared. If the gravity becomes larger, the position information after the replacement is replaced and retained. S7. Set the number of cycles, and iterate steps S5-S6 with the number of cycles to obtain the task allocation strategy with the minimum total overhead.

2. A method for task allocation in a mobile edge computing network considering task delay and server cost as claimed in claim 1, characterized in that: The computational delay of each task in the task set in the edge computing server and the mobile cloud server is calculated by dividing the computational amount of each task by the number of CPU cycles that the edge computing server and the mobile cloud server can provide per second.

3. A method for task allocation in a mobile edge computing network considering task delay and server cost as claimed in claim 1, characterized in that: If two tasks in the task set that have a successor relationship are on the same server, the transmission delay is 0.

4. A method for allocating tasks in a mobile edge computing network taking into account task delay and server cost as claimed in claim 1, characterized in that: In step S2, the task set determines the energy consumption of the server in combination with the hardware architecture of the server when the user local server, the edge computing server, and the mobile cloud server perform energy consumption calculations.

5. A method for task allocation in a mobile edge computing network considering task delay and server cost as claimed in claim 1, characterized in that: The average calculation delay of the task processed on the server in step S3 is obtained by adding the upload delay of the task, the calculation delay of the task on the server and the download delay after the task calculation is completed.

6. A method for task allocation in a mobile edge computing network considering task delay and server cost as claimed in claim 1, characterized in that: The server execution cost is calculated based on the computing delay of the task on the server, the billing cycle of the server, the computing power of the server, the minimum server computing power and the basic price of the server.

Images