WO2019200716A1

WO2019200716A1 - Fog computing-oriented node computing task scheduling method and device thereof

Info

Publication number: WO2019200716A1
Application number: PCT/CN2018/093934
Authority: WO
Inventors: 王昆仑; 杨旸; 李强; 周明拓
Original assignee: 上海无线通信研究中心; 福州物联网开放实验室有限公司
Priority date: 2018-04-20
Filing date: 2018-06-30
Publication date: 2019-10-24
Also published as: CN110392079A; CN110392079B

Abstract

Disclosed by the present invention is a fog computing-oriented node computing task scheduling method and device thereof. The method comprises: computing energy efficiency when task computing and task migration are performed between a user equipment and a neighboring equipment; obtaining, by means of the mathematical optimization method, an optimal energy efficiency communication time of the task migration between the user equipment and the neighboring equipment and a corresponding task migration amount; and migrating, by the user equipment undertaking the computing task, the computing task to an idle neighboring equipment for computing according to the optimal energy efficiency communication time and the corresponding task migration amount. According to the present invention, an idle equipment around the user equipment can implement optimal migration of computing tasks, thereby sharing the computing resources thereof, so that resource-constrained user equipment migrates excessive computing tasks to nearby idle equipment with sufficient resources, thus reducing service delay to the greatest extent and achieving high energy efficiency of communication.

Description

Node computing task scheduling method and device for fog calculation

Technical field

The invention relates to a node computing task scheduling method for fog calculation, and relates to a fog computing device for implementing the node computing task scheduling method, which belongs to the technical field of computing communication.

Background technique

With the continuous evolution of network architecture, advanced distributed computing concepts such as cloud computing and fog computing have been proposed to cope with the explosive growth of data traffic demand and low latency business challenges. Cloud Computing centralizes storage, control, and processing power in powerful cloud servers, leveraging centralized data processing and resource management to improve network resource utilization efficiency and energy efficiency. Fog Computing is a cloud computing that is scattered around people. In fog computing, data, data processing, and applications are concentrated in network edge devices, rather than being stored almost entirely in cloud servers.

Compared with cloud computing, fog computing relies not on cloud servers in a central location, but on distributed computer resources that are closer to local devices. As shown in FIG. 1 , a typical fog computing network system utilizes functions of computing, storage, communication, management, and the like from a cloud server to a network edge device to a user device, forming a continuous service area from the cloud server to the terminal. The network edge device may be a traditional network device, such as a base station, a router, a switch, a gateway, etc., which has been deployed in the network, or a dedicated local server. Fog computing brings the boundaries of the cloud closer to the "edge" of the local server (connected to the Internet of Things), rather than keeping them almost entirely in the cloud like cloud computing. It can be seen that unlike cloud computing floating at the far end, fog computing is to let the computing unit be dispersed around people like fog, so as to achieve the most efficient use of computing resources.

In the fog computing network system, direct communication between devices and devices can be realized without the help of network devices such as base stations, which can improve the spectral efficiency of data transmission and achieve efficient load balancing. Based on these advantages, direct communication between devices will occupy a major position in next-generation wireless communications. On the other hand, mobile applications are becoming more and more complex, and because of the low latency requirements of these complex applications, user equipment requires a large amount of computing and communication resources to ensure real-time performance. Therefore, migrating the computing tasks of the user equipment to the adjacent idle devices can realize resource sharing and ensure efficient operation with low latency, and has a good application prospect.

Summary of the invention

The primary technical problem to be solved by the present invention is to provide a node computing task scheduling method for fog calculation.

Another technical problem to be solved by the present invention is to provide a fog computing device that implements the above-described node computing task scheduling method.

In order to achieve the above object, the present invention adopts the following technical solutions:

According to a first aspect of the embodiments of the present invention, a node computing task scheduling method for fog calculation is provided, including the following steps:

Calculate the energy efficiency when performing task calculation and task migration between the user equipment and the neighboring equipment;

Through the mathematical optimization method, the optimal energy-efficient communication time of the task migration between the user equipment and the neighboring equipment and the corresponding task migration amount are obtained;

The user equipment undertaking the computing task migrates the computing task to the idle neighboring device for calculation according to the optimal energy-efficient communication time and the corresponding task migration amount.

Preferably, the user equipment and the neighboring device multiplex the spectrum of the cellular network by using an opportunistic spectrum sharing mechanism to implement direct communication between devices.

Preferably, the user equipment migrates the computing task to the neighboring device by using a time division multiplexing manner or an orthogonal frequency division multiple access method.

Preferably, the energy efficiency is calculated by the following formula:

Among them, u _ee is energy efficiency, l is the task migration amount of the user equipment, E _i is the energy consumption of the task migration, E _re is the energy consumption calculated by the task, E ₀ is the circuit energy consumption, and K is a positive integer.

Preferably, the task migration amount

among them

Indicates the probability of successfully accessing the jth spectrum resource,

For the average transmission time obtained in the time slot T for the direct communication between the i-th pair of devices, W and b _i,j respectively represent the modulation mode of the bandwidth and the transmission, and i and j are positive integers.

Preferably, the energy consumption of the task migration

among them

Indicates the probability of successfully accessing the jth spectrum resource,

The average transmission time obtained in the time slot T for the direct communication between the i-th pair of devices, f(b _i,j ) is the power expression of the task migration, and i and j are positive integers.

Which is better,

Where σ ² is the noise power, ε task migration uses the symbol error probability when the modulation mode is b _i,j , and φ(d _i,i ) represents the channel fading parameter when the distance between devices is d _i,i .

According to a second aspect of the embodiments of the present invention, there is provided a fog computing device comprising a processor and a memory, the processor reading a computer program in the memory for performing the following operations:

Calculating energy efficiency when task calculation and task migration are performed between the fog computing device and the adjacent fog computing device;

Through the mathematical optimization method, the optimal energy-efficient communication time and the corresponding task migration amount of the task migration of the fog computing device and the adjacent fog computing device are obtained;

The fog computing device undertaking the computing task migrates the computing task to the adjacent idle fog computing device for calculation according to the optimal energy efficiency communication time and the corresponding task migration amount.

Preferably, the fog computing device and the adjacent fog computing device multiplex the spectrum of the cellular network through an opportunistic spectrum sharing mechanism to implement direct communication between devices.

Preferably, the fog computing device migrates the computing task to the adjacent fog computing device by means of time division multiplexing or orthogonal frequency division multiple access.

Compared with the prior art, the node computing task scheduling method and device provided by the present invention can process as many computing tasks as possible with as little migration energy as possible. With the present invention, user equipments that are idle around the user equipment (referred to as idle devices) can implement optimal migration of computing tasks, thereby sharing their computing resources (CPU, GPU, etc.), so that resource-constrained user equipments will be excessively calculated. Tasks (online games, virtual reality simulation, etc.) are migrated to nearby idle devices with sufficient resources to minimize business latency and achieve high energy efficiency of communication.

DRAWINGS

1 is a diagram showing an example of a working scenario of a typical fog computing network system;

2 is a diagram showing an example of a fog computing network system for implementing the present invention;

FIG. 3 is a flowchart of an operation for implementing a task task uninstallation according to an embodiment of the present invention; FIG.

Figure 4 is a simulation comparison diagram of performance comparison of different computing task migration methods;

Fig. 5 is a view showing an example of the structure of a fog calculating apparatus for carrying out the present invention.

detailed description

The technical content of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

In the fog computing network system, the computing resources are disclosed to the undirected users accessing the local network, who needs to use the application, increase the resource utilization by sharing, and improve the reliability of the entire network system in a redundant manner. In this kind of computing resource allocation mode, for the task-intensive users, if the computing task is migrated to the cloud server, the power consumption of the entire network system and the delay of data access are relatively large. On the other hand, the migration of computing tasks to the cloud server imposes a heavy burden on the server, which affects the user's need for business access. The migration of computing tasks to nearby small base stations equipped with servers also has similar problems for remote users.

To this end, the node computing task scheduling method provided by the present invention firstly studies the energy consumption of the task calculation and task migration of the fog computing network system, and then uses the mathematical optimization method to solve the overall energy efficiency optimization problem according to the system constraints, and finally according to the mathematics. The optimized solution obtains the optimal energy-efficient communication time and the optimal task amount to be migrated to each adjacent device when the user equipment is paired with each adjacent user equipment (referred to as an adjacent device).

The detailed description of the above-mentioned node calculation task scheduling method will be described below.

First, some new technical concepts and their parameters used in the present invention are introduced in conjunction with Table 1.

Table 1 Parameter comparison table

It is assumed that the user equipment S as the fog computing device needs to calculate the task amount in a time slot as R _S , wherein the l-bit task quantity needs to be migrated to the neighboring device, and the other (R _S -l)-bit passes. The calculation is performed by its own central processing unit (CPU). In general, most of the calculated results of the task are small relative to the task itself. Here, it can be assumed that the time and energy cost of the neighboring device to feed back the calculation result to the user equipment is negligible.

Let the set of spectrum resource blocks of the cellular network be M={1, 2, . . . , N _m }, and the user equipment migrates the computing task to K (K is a positive integer) neighboring devices by time division multiplexing, and any device communicates with the device. Group i can randomly allocate time resources τ _i , i ∈ {1, 2, ..., K} of any size. Let the power consumption of the i-th pair device and device communication group use the jth resource block for task migration is P _i,j and the noise power is σ ² .

It is assumed that all of the fog computing devices have the same circuit power consumption P ₀ , correspondingly the circuit energy consumed at the migration time T is E ₀ . Similar to the traditional energy efficiency definition, the energy efficiency here takes into account the energy consumption E _off when the task is migrated and the energy consumption E _comp when the task is calculated. Therefore, the energy efficiency in the present invention can be defined as the task amount with the smallest energy consumption migration. Considering the task amount of the user equipment migration is 1, the energy efficiency can be expressed as:

On the other hand, in the fog computing network system, common methods for calculating task migration include the following two types:

Migrate computing tasks to the cloud server

Migrate computing tasks to adjacent small base stations

Compared with the task migration of the traditional cloud server or the base station, the task migration based on the direct communication between the devices can perform the optimal energy-efficient task migration according to the computing resources of the neighboring devices and the calculation energy consumption, and ensure the optimal task amount for each adjacent device to migrate. . Therefore, the key to the successful application of the task scheduling method based on the optimal energy efficiency is how to solve the optimal solution problem based on the energy efficiency expression.

In order to solve the above problem, in the fog computing network system shown in FIG. 2, it is assumed that the spectrum resource used for direct communication between devices is through an opportunistic spectrum sharing mechanism (ie, when the cellular system does not occupy a certain spectrum resource, the fog computing network system) Reusing the idle spectrum resource of the segment. The spectrum sharing mechanism mainly accesses the spectrum resource of the cellular communication by means of the monitoring of the spectrum by the fog computing device, and according to the definition, the direct communication between the device of the i-th pair can be obtained in the time slot. The average transmission time obtained in T is

among them

Indicates the probability of successfully accessing the jth spectrum resource. Correspondingly, the amount of tasks that can be migrated between the i-th pair of devices can be obtained as

Where W and b _i,j represent the modulation of the bandwidth and transmission _, respectively. When the task migration adopts the modulation mode of b _i,j , the symbol error probability is

Where χ _i,j represents the received signal to noise ratio. According to the working principle of the opportunistic spectrum sharing mechanism, the expression of the received signal to noise ratio of the i-th pair of devices can be obtained as

Where φ(d _i,i ) represents the channel fading parameter when the distance between devices is d _i,i .

According to the function relationship, the approximate solution of the symbol error rate and the power expression of the task migration are obtained.

Substituting the spectrum sharing power and the time resource allocation, the energy consumption of the task migration of the i-th pair of inter-device communication groups can be obtained as

among them

Since the computing task can be migrated to K neighboring devices, the total task migration energy consumption can be expressed as

"Efficient multi-user computation offloading for mobile-edge cloud computing" (published in IEEE/ACM Trans. Networking, vol.), co-published by X. Chen, L. Jiao, W. Li, and X.Fu. 24, no. 5, pp. 2795-2808, Oct. 2016), the technical idea of using the communication between the device and the device to migrate the computing task is proposed, but the paper does not fully consider the problem of energy efficiency. In the present invention, Referring to the technical idea provided by the paper, it is assumed that the CPU cycle required for the i-th neighboring device to calculate the 1-bit task is C _re,i , and the energy consumed for the calculation of one cycle is P _re,i . Then C _re,i P _re,i can be expressed as the energy consumed by the 1-bit calculation. Assuming that the task quantity required to be calculated by the i-th neighboring device is l _i -bit, the calculated energy consumed by the device can be expressed as E _re,i , According to the previous derivation, you can get:

Here, it can be assumed that the CPU frequencies of different neighboring devices can be different, and the CPU frequency of the same device is fixed.

According to the energy consumption closed expression and the definition of energy efficiency, which are derived from the above task calculation and task migration, the energy efficiency expression of the entire fog computing network system can be obtained during the migration computing task:

In order to achieve the optimal energy efficiency of the entire fog computing network system, the required task migration can be obtained through calculation. Therefore, according to the system constraints, the optimization problem that needs to be solved is:

St b _i,j ∈[b _min ,b _max ],

C _re,i l _i ≤C _i ,

l _i ≥ 0,

τ _i ≥0.

Transform the above optimization problem into an equivalent low complexity problem

among them

Through the mathematical optimization method, the optimal energy-efficient communication time for task migration of the direct communication group i between each pair of devices can be obtained.

And the corresponding task migration

Finally, the fog computing device that undertakes the computing task is based on the optimal energy efficiency communication time.

And the corresponding task migration

Migrate computing tasks to idle neighbors for calculation. When task migration is required, the data transmitting end uses the optimal energy-efficient modulation mode for data transmission. The so-called optimal energy efficiency modulation method refers to the number of bits containing the optimal energy efficiency in the symbols of data transmission, that is, b _i,j for solving the optimization problem.

The node computing task scheduling method provided by the present invention is described below in conjunction with a specific embodiment. In this embodiment, the mission device S and the neighboring device B in the fog computing network system are capable of pairing for communication. The task device S needs to perform a computing task uninstallation. Since direct communication between these devices does not have its own spectrum resources, it is necessary to multiplex the spectrum resources of the cellular network. Therefore, it is assumed that spectrum resources in a cellular network are used to implement direct communication between devices, and the spectrum of the cellular network is multiplexed by an opportunistic spectrum sharing mechanism. Referring to FIG. 3, the specific steps for implementing the uninstallation of the computing task in this embodiment are as follows:

Step A1: According to the total amount of tasks to be migrated and the energy of the migration, calculate the energy consumption, and obtain an energy efficiency expression:

Step A2: Energy efficiency optimization problem of computing task migration according to system constraints

Step A3: According to the form of the energy-efficient closed expression, the above optimization problem is equivalent to the low complexity optimization problem

Step A4: Obtain optimal time resource allocation by mathematical optimization method

And task migration

Step A5: It is assumed that the task device S has K neighboring devices, and the task device S passes the TDMA (Time Division Multiple Access) mode or the orthogonal frequency division multiple access mode (which requires combining multiple antenna technologies), based on the above optimal Time resource allocation

And task migration

Migrate computing tasks to these K neighboring devices.

In the process of solving the optimal solution of the algorithm, the optimal calculation method of task migration can be obtained by a low complexity algorithm. Specifically, the original optimization problem is first translated into the following questions:

among them,

Then, it is proved that the above optimization problem is equivalent to solving the following problems:

Finally, the optimal solution can be solved by the iterative convex optimization method with low computational complexity.

Figure 4 shows the performance comparison simulation results of different computational task migration methods in terms of energy efficiency. It can be seen from the comparison of the simulation results that the computational task migration scheme of the optimal energy efficiency (including optimal time allocation and optimal migration rate) provided by the present invention is always superior to the conventional scheme in performance. Moreover, under the same time resource allocation conditions, selecting the optimal energy efficiency transmission rate for task migration will have a great performance gain.

Further, the present invention also provides a fog computing device that implements the above node computing task scheduling method. As shown in FIG. 5, the fog computing device includes at least a processor and a memory, and further includes a communication component, a sensor component, a power component, a multimedia component, and an input/output interface according to actual needs. The memory, the communication component, the sensor component, the power component, the multimedia component, and the input/output interface are all connected to the processor. In an embodiment of the invention, the memory in the fog computing device may be a static random access memory (SRAM), an electrically erasable programmable read only memory (EEPROM), an erasable programmable read only memory (EPROM), Programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, etc., the processor can be a central processing unit (CPU), graphics processing unit (GPU), field programmable logic gate array (FPGA) ), application specific integrated circuits (ASIC), digital signal processing (DSP) chips, and the like. Other communication components, sensor components, power components, multimedia components, and the like can be implemented by using common components in existing smart phones, and are not specifically described herein.

In the above-mentioned fog computing device, the processor reads a computer program in the memory for performing the following operations: calculating energy consumption when performing task calculation and task migration between the user equipment and the adjacent device; obtaining by mathematical optimization method The optimal energy-efficient communication time of the user equipment and the adjacent equipment for task migration and the corresponding task migration amount; the user equipment undertaking the calculation task migrates the calculation task to the idle neighboring device according to the optimal energy-efficient communication time and the corresponding task migration amount. Calculation. As one of the preferred solutions, the user equipment and the neighboring device multiplex the spectrum of the cellular network through an opportunistic spectrum sharing mechanism to implement direct communication between devices. As a second preferred solution, the user equipment migrates the computing task to the neighboring device by means of time division multiplexing.

Compared with the prior art, the present invention considers an optimal computing task migration scheme from the perspective of resource utilization, and uses the direct communication between devices to migrate computing tasks while considering how to allocate these tasks to the optimal energy efficiency. Different proximity devices and solved the optimal task migration. With the present invention, idle devices around the user equipment can achieve optimal migration of computing tasks, thereby sharing their computing resources (CPU, GPU, etc.), so that resource-constrained user equipment will have too many computing tasks (online games, virtual reality) Simulation, etc.) migrate to idle devices with sufficient resources nearby, minimizing service delay and achieving high energy efficiency of communication.

The node calculation task scheduling method and device for fog calculation provided by the present invention are described in detail above. Any obvious changes made to the invention without departing from the spirit of the invention will constitute an infringement of the patent right of the present invention and will bear corresponding legal liabilities.

Claims

A node computing task scheduling method for fog computing, comprising the following steps:

Calculate the energy efficiency when performing task calculation and task migration between the user equipment and the neighboring equipment;

Through the mathematical optimization method, the optimal energy-efficient communication time of the task migration between the user equipment and the neighboring equipment and the corresponding task migration amount are obtained;

The user equipment undertaking the computing task migrates the computing task to the idle neighboring device for calculation according to the optimal energy-efficient communication time and the corresponding task migration amount.
The node computing task scheduling method according to claim 1, wherein:

The user equipment and the neighboring device multiplex the spectrum of the cellular network through an opportunistic spectrum sharing mechanism to implement direct communication between devices.
The node computing task scheduling method according to claim 1, wherein:

The user equipment migrates the computing task to the neighboring device by means of time division multiplexing or orthogonal frequency division multiple access.
The node computing task scheduling method according to claim 1, wherein the energy efficiency is calculated by the following formula:

Among them, u ee is energy efficiency, l is the task migration amount of the user equipment, E i is the energy consumption of the task migration, E re is the energy consumption calculated by the task, E 0 is the circuit energy consumption, and K is a positive integer.
The node computing task scheduling method according to claim 4, wherein:

Task migration
among them
Indicates the probability of successfully accessing the jth spectrum resource,
For the average transmission time obtained in the time slot T for the direct communication between the i-th pair of devices, W and b i,j respectively represent the modulation mode of the bandwidth and the transmission, and i and j are positive integers.
The node computing task scheduling method according to claim 4, wherein:

Energy consumption of the task migration
among them
Indicates the probability of successfully accessing the jth spectrum resource,
The average transmission time obtained in the time slot T for the direct communication between the i-th pair of devices, f(b i,j ) is the power expression of the task migration, and i and j are positive integers.
The node computing task scheduling method according to claim 6, wherein:

Where σ 2 is the noise power, ε task migration uses the symbol error probability when the modulation mode is b i,j , and φ(d i,i ) represents the channel fading parameter when the distance between devices is d i,i .
A fog computing device is characterized by comprising a processor and a memory, the processor reading a computer program in the memory for performing the following operations:

Calculating energy efficiency when task calculation and task migration are performed between the fog computing device and the adjacent fog computing device;

Through the mathematical optimization method, the optimal energy-efficient communication time and the corresponding task migration amount of the task migration of the fog computing device and the adjacent fog computing device are obtained;

The fog computing device undertaking the computing task migrates the computing task to the adjacent idle fog computing device for calculation according to the optimal energy efficiency communication time and the corresponding task migration amount.
The fog computing device of claim 8 wherein:

The fog computing device and the adjacent fog computing device multiplex the spectrum of the cellular network through an opportunistic spectrum sharing mechanism to implement direct communication between devices.
The fog computing device of claim 8 wherein:

The fog computing device migrates the computing task to the adjacent fog computing device by means of time division multiplexing or orthogonal frequency division multiple access.