A task offloading and resource allocation method with energy consumption and latency trade-off in mobile edge computing networks

Abstract

The present invention relates to a method for task unloading and resource allocation with a compromise between energy consumption and delay in a mobile edge computing network, the method comprising the steps of: S1, obtaining network configuration information; S2, taking the joint minimization of energy consumption and average computing delay as the goal, generating two sub-goals of unloading decision and resource allocation decision; S3, fixing the resource allocation decision, and determining the unloading decision with the minimum energy consumption and average computing delay; S4, fixing the unloading decision, and calculating the resource allocation strategy with the minimum energy consumption and average computing delay; S5, looping and iterating S3-S4, obtaining the unloading decision and resource allocation decision with the minimum energy consumption and total delay overhead; S6, determining the task unloading and resource allocation of the mobile edge computing network. The system of the present invention decomposes the energy consumption and average computing delay optimization goals into two unloading decision and resource allocation decision sub-problems, so that the minimum value can be iteratively determined, and the average computing overhead of users is minimized.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a task offloading and resource allocation method that compromises energy consumption and delay in a mobile edge computing network. Background Art

[0002] With the popularization of large-scale mobile terminal devices, high-speed or ultra-low latency data traffic has exploded. Mobile devices with limited battery capacity and computing power cannot meet the growing user needs. A new solution is to offload computing tasks to mobile edge computing (MEC) servers to improve system performance. Compared with the cloud with powerful computing power, mobile edge computing servers usually have limited computing resources and cannot meet the computing needs of all users at the same time. Based on this, how to achieve reasonable allocation of resources has become a key issue that needs to be solved in cloud fog networks. Summary of the invention

[0003] Based on the above-mentioned shortcomings and deficiencies in the prior art, one of the objects of the present invention is to solve at least 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 compromises energy consumption and latency and meets one or more of the above-mentioned requirements.

[0004] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:

[0005] A task offloading and resource allocation method for mobile edge computing network with a trade-off between energy consumption and latency, the method comprising the steps of:

[0006] S1. Obtain network configuration information, where the network configuration information includes task computing capacity and local computing capacity of each user in the network;

[0007] S2, with the goal of minimizing the total overhead of energy consumption and average computing delay, generating two sub-goals of offloading decision and resource allocation decision according to the computing task and the network configuration information;

[0008] S3, fix the resource allocation decision, compare the energy consumption and average computing delay of the task processed locally with the energy consumption and average computing delay of the task processed on the server, and determine the offloading decision with the minimum total overhead of energy consumption and average computing delay;

[0009] S4, fixing the offloading decision, and calculating a resource allocation strategy with minimum energy consumption and average computing delay according to the system bandwidth and resource allocation situation;

[0010] S5. Set the number of cycles, and iterate steps S3-S4 with the number of cycles to obtain the unloading decision and resource allocation decision with the minimum energy consumption and total delay overhead;

[0011] S6. Determine task offloading and resource allocation of a mobile edge computing network according to the offloading decision and the resource allocation decision.

[0012] As a preferred solution, the local computing power of each user is determined according to the number of CPU cycles that each user can provide per second.

[0013] As a further preferred solution, the task calculation amount is determined by multiplying the task size and the number of CPU cycles required to process each bit in the task.

[0014] As a preferred solution, in step S2, the energy consumption of the server is determined in combination with the hardware architecture of the mobile edge computing network when calculating the energy consumption.

[0015] As a preferred solution, minimizing the total overhead of the energy consumption and the average computing delay is to set a weight for the energy consumption and the average computing delay respectively, and obtain the total computing overhead by calculating the weights of the energy consumption and the average computing delay.

[0016] As a preferred solution, 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.

[0017] As a preferred solution, the energy consumption of the task in the server processing in step S3 is obtained according to the energy consumption on the user side during the task uploading process.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] The method of the present invention decomposes the energy consumption and average computing delay optimization objectives of the mobile edge computing network into two offloading decision and resource allocation decision sub-problems, so that the minimum values ​​of energy consumption and average computing delay can be iteratively determined, thereby minimizing the average computing overhead of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flowchart of a task offloading and resource allocation method that compromises energy consumption and latency in a mobile edge computing network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] 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.

[0022] Embodiment: This embodiment provides a task offloading and resource allocation method that compromises energy consumption and latency in a mobile edge computing network, and its flow chart is as follows: Figure 1 As shown: First, step S1 is performed to obtain network configuration information, which includes the task computing amount and the local computing power of each user in the network; in this embodiment, the task computing amount can be expressed as T k (D k ,C k );where D k Indicates the size of the computing task, C k Represents the number of CPU cycles required to process each bit in the task. The local computing power of each user in the network can be expressed as Represents user u k The local processing power of a device is the number of CPU cycles it can provide per second.

[0023] After obtaining the above network configuration information, step S2 can be started, with the goal of minimizing energy consumption and average computing delay, generating two sub-problems of offloading decision and resource allocation decision according to the network configuration information;

[0024] Specifically, step S2 in this embodiment may include the following steps:

[0025] Construct a set L = {l 1 ,l 2 ,L l k ,L,l K} represents the computation offloading strategy of user set U, where l k Represents user u k The uninstallation decision is as follows:

[0026]

[0027] Concurrent Build Represents the set of policies that MEC uses to allocate computing resources to offloading users, where Assigned to user u by the MEC server k The calculation rate, F MEC Indicates the maximum number of CPU cycles that the server can provide per second, satisfying

[0028] Then proceed to step S3, fix the resource allocation decision, compare the energy consumption and average computing delay of the task processed locally with the energy consumption and average computing delay of the task processed on the server, and determine the offloading decision with the minimum energy consumption and average computing delay;

[0029] First, the total overhead of local processing of tasks by each user is calculated, that is, the average computing delay and energy consumption.

[0030] According to the local processing capacity of each user obtained in step S1, the delay of local processing calculation of the task is expressed as

[0031] Among them, D k Indicates the size of the computing task, C k Indicates the number of CPU cycles required to process each bit in the task.

[0032] Furthermore, the energy consumption calculation formula of the task in local calculation is listed:

[0033]

[0034] The above k n A coefficient indicating how the computing speed varies according to the hardware architecture;

[0035] Combined with the above local processing and calculation delay and local computing energy consumption Calculate user u k The total cost of processing the task locally is:

[0036]

[0037] In the above formula, and Represents user u k The weight of computing latency and energy consumption in the total computing overhead.

[0038] Next, the total delay of the user uploading the task to the mobile edge computing network for calculation is calculated. The total delay of the task offloading calculation in this embodiment can be expressed as:

[0039]

[0040] In the above formula, Indicates the upload time of the task. represents the computing time of the mobile edge computing network server, Indicates the download time after the task calculation is completed.

[0041] In order to calculate the total delay, a set α = {α 1 ,α 2 ,L,α K} represents the system bandwidth allocation strategy, α k ∈[0,1] represents the normalized proportional coefficient of the system bandwidth, satisfying

[0042] Assume user u k The task upload rate is Where P kRepresents user u k The transmission power, h km Represents user u k Uplink status of task upload to BS, n 0 Represents the power of additive white Gaussian noise.

[0043] Further assume that user u k The task upload time is User k The computation time at the mobile edge computing network server is After the offloading task is completed, the download rate is In the above formula, P M Indicates the transmit power of the BS.

[0044] From this, we can calculate the output-to-input ratio of the task as β k , then the calculated data size is β k D k , then user u k The download time of the task at the MEC server is

[0045] Combining the formulas set above, we can get the calculation delay of the task processing on the server as:

[0046]

[0047] For mobile edge networks, since their servers are generally actively connected, the energy consumption of task offloading calculations only needs to consider the energy consumption on the client side during task uploading:

[0048]

[0049] Combining the computational latency and energy consumption of the above tasks on the server, the total computational overhead of task offloading can be obtained as and Represents user u k The weight of computing latency and energy consumption in the total computing overhead.

[0050] For user u k For example, when the uninstall decision is l k In the case of , the total cost of completing the computing task is expressed as:

[0051]

[0052] Since the task attributes and device status on the user side can be known in advance, the cost of performing local calculations on the task is low. is a known constant, so the calculation of minimizing the energy consumption and average computing delay of tasks processed locally and processed on the server is summarized into the following six formulas:

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] Since the above calculation of minimizing the energy consumption and average computing delay of tasks processed locally with the energy consumption and average computing delay of tasks processed on the server is a mixed integer type problem, it is difficult to obtain the optimal solution. Therefore, it is decomposed into two sub-problems, namely, offloading decision and resource allocation decision.

[0060] When both the bandwidth allocation ratio coefficient α and the computing resource allocation strategy F are determined, the offloading decision subproblem can be rewritten as follows:

[0061]

[0062]

[0063] According to the above formula, in order to minimize the average user computing overhead, for each user, if the task is to be uploaded to the server of the mobile edge computing network for processing under a given resource allocation strategy, it must satisfy

[0064] When uninstalling k =1, so the user's optimal uninstall decision is L * = {l 1 * ,L,l k * ,L,l K *},in

[0065] When the user's offloading decision L is solved, assuming that the set of users who choose to offload tasks is U' and the total number of offloading users is K', the resource allocation subproblem can be reformulated as:

[0066]

[0067]

[0068]

[0069]

[0070]

[0071] because It is only related to the local device status and the characteristics of its own computing task, and can be determined in advance before the calculation, so it is a known constant. Therefore, for the resource allocation sub-problem, the optimization objective can be rewritten as:

[0072] This formula is used to determine the offloading decision with the minimum energy consumption and average computing delay.

[0073] After the unloading decision with the minimum energy consumption and average computing delay is determined, step S4 is performed to fix the unloading decision and calculate the resource allocation strategy with the minimum energy consumption and average computing delay according to the system bandwidth and resource allocation situation;

[0074] make Then there is in

[0075] According to the above formula, we can know In addition, when k≠m,

[0076] Similarly, we can get: Due to the non-negativity of the parameters in the formula, we can get In addition, when k≠m,

[0077] According to

[0078] Get the second-order Hessian matrix H of f(α,F):

[0079]

[0080] From the above analysis, we can see that the second-order matrix H is positive definite, and because the constraints are linear, the optimization of the minimum resource allocation strategy is a convex optimization problem.

[0081] For this optimization problem, we first construct the Lagrangian function to solve it. In the above formula, λ and μ represent Lagrange multipliers, and λ>0,μ>0.

[0082] The corresponding dual function is

[0083] Based on the characteristics of the minimum resource allocation strategy optimization being a convex optimization problem, the minimum value can be obtained only when the first-order derivative of α,F is 0. The first-order derivatives are:

[0084]

[0085]

[0086] make have to:

[0087]

[0088] α k * and Substituting into the Lagrangian function, we get the dual function:

[0089]

[0090] The dual problem is The first-order derivative of the dual function with respect to λ and μ is:

[0091]

[0092]

[0093] make have to λ * , μ * Substitute α respectively k * and At this time, the best resource allocation strategy is:

[0094]

[0095]

[0096] In this way, an offloading and resource allocation method with minimal energy consumption and average computing delay is obtained.

[0097] In order to further reduce the energy consumption and the average calculation delay, step S5 is performed, the number of loops N is set, and steps S3 and S4 are looped N times to iteratively gradually reduce the energy consumption and the average calculation delay to the minimum value.

[0098] After obtaining the task offloading and resource allocation decision with the minimum energy consumption and average computing delay through the above method, task offloading and resource allocation of the mobile edge computing network can be carried out according to the task offloading and resource allocation decision.

[0099] 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 offloading and resource allocation method with a trade-off between energy consumption and latency in a mobile edge computing network, characterized in that: The method comprises the steps of: S1. Obtain network configuration information, which includes task computing amount and local computing power of each user in the network; task computing amount is represented by T k (D k ,C k ), where D k Indicates the size of the computing task, C k Represents the number of CPU cycles required to process each bit in the task; the local computing power of each user in the network is expressed as Represents user u k The local processing power, which represents the number of CPU cycles provided by the device per second; S2, with the goal of minimizing the total cost of energy consumption and average computing delay, generating two sub-goals of offloading decision and resource allocation decision according to the computing task and the network configuration information, specifically including the following steps: Construct a set L = {l1,l2,…l k ,…,l K } represents the computation offloading strategy of user set U, where l k Represents user u k The uninstallation decision is as follows: Concurrent Build Represents the set of policies that MEC uses to allocate computing resources to offloading users, where Assigned to user u by the MEC server k The calculation rate, F MEC Indicates the maximum number of CPU cycles per second provided by the server, satisfying S3, fixing the resource allocation decision, comparing the energy consumption and average computing delay of the task processed locally with the energy consumption and average computing delay of the task processed on the server, and determining the offloading decision with the minimum energy consumption and average computing delay, specifically including the following steps: The total cost of computing tasks processed locally by each user, that is, the average computing latency and energy consumption; According to the local processing capacity of each user obtained in step S1, the delay of local processing calculation of the task is expressed as Among them, D k Indicates the size of the computing task, C k Indicates the number of CPU cycles required to process each bit in the task; Furthermore, the energy consumption calculation formula of the task in local calculation is listed: The above k n A coefficient indicating how the computing speed varies according to the hardware architecture; Combined with the above local processing and calculation delay and local computing energy consumption Calculate user u k The total cost of processing the task locally is: In the above formula, and Represents user u k The weight of computing latency and energy consumption in the total computing overhead; Then the total delay of the user uploading the task to the mobile edge computing network for calculation is calculated. The total delay of task offloading calculation is expressed as: In the above formula, Indicates the upload time of the task. represents the computing time of the mobile edge computing network server, Indicates the download time after the task calculation is completed; Set a set α={α1,α2,…,α K } represents the system bandwidth allocation strategy, α k ∈[0,1] represents the normalized proportional coefficient of the system bandwidth, satisfying Assume user u k The task upload rate is Where P k Represents user u k The transmission power, h km Represents user u k The uplink status of the task uploaded to the BS, n0 represents the power of the additive Gaussian white noise; Further assume that user u k The task upload time is User k The computation time at the mobile edge computing network server is After the offloading task is completed, the download rate is In the above formula, P M Indicates the transmit power of the BS; The output-to-input ratio of the calculation task is β k , then the calculated data size is β k D k , then user u k The download time of the task at the MEC server is Combining the formulas set above, we can get the calculation delay of the task processing on the server as: Further consider the energy consumption on the client side during task upload: Combining the computational latency and energy consumption of the above tasks on the server, the total computational overhead of task offloading is: and Represents user u k The weight of computing latency and energy consumption in the total computing overhead; For user u k For example, when the uninstall decision is l k In the case of , the total cost of completing the computing task is expressed as: The calculation of minimizing the energy consumption and average computing delay of tasks processed locally and processed on the server is summarized into the following six formulas: When both the bandwidth allocation ratio coefficient α and the computing resource allocation strategy F are determined, the offloading decision subproblem is rewritten as follows: According to the above formula, in order to minimize the average user computing overhead, for each user, if the task is to be uploaded to the server of the mobile edge computing network for processing under a given resource allocation strategy, it must satisfy When uninstalling k =1, so the user's optimal uninstall decision is L * = {l1 * ,…,l k * ,…,l K * },in When the user's offloading decision L is solved, assuming that the set of users who choose to offload tasks is U' and the total number of offloading users is K', the resource allocation subproblem is reformulated as: For the resource allocation subproblem, the optimization objective is rewritten as: This formula is used to determine the offloading decision with the minimum energy consumption and average computing delay; S4, fixing the offloading decision, and calculating the resource allocation strategy with the minimum total overhead of energy consumption and average computing delay according to the system bandwidth and resource allocation situation, which specifically includes the following steps: make Then there is in According to the above formula, we can know In addition, when k≠m, Similarly, we can get: Due to the non-negativity of the parameters in the formula, we can get In addition, when k≠m, According to Get the second-order Hessian matrix H of f(α,F): Reconstruct the Lagrangian function to solve In the above formula, λ and μ represent Lagrange multipliers, and λ>0,μ>0; The corresponding dual function is Based on the characteristics of the minimum resource allocation strategy optimization being a convex optimization problem, the minimum value can be obtained only when the first-order derivative of α,F is 0. The first-order derivatives are: make have to: α k * and Substituting into the Lagrangian function, we get the dual function: The dual problem is The first-order derivative of the dual function with respect to λ and μ is: make have to λ * , μ * Substitute α respectively k * and At this time, the best resource allocation strategy is: In this way, the offloading and resource allocation method with the minimum energy consumption and average computing delay is obtained; S5. Set the number of cycles, and iterate steps S3-S4 with the number of cycles to obtain the unloading decision and resource allocation decision with the minimum energy consumption and total delay overhead; S6. Determine task offloading and resource allocation of a mobile edge computing network according to the offloading decision and the resource allocation decision.

Images