CN109413623B - Cooperative computing migration method between energy-starved terminal and flow-starved terminal - Google Patents

Abstract

The invention discloses a cooperative calculation migration method between an energy-starved terminal and a flow-starved terminal, which utilizes the advantages of the flow-starved terminal and the energy-starved terminal to make up the respective inferior resources and effectively calculate and migrate on the premise of not exhausting the electric quantity of a battery or generating extra flow cost. The method comprises the following steps: step 10) the initiator broadcasts the demand signaling over the control channel, when the helper responds to the initiator, collaboration starts, ES_iObtaining DS through control channel_jTask parameter of, DS_jObtaining ES through control channel_iThe task parameters of (1); step 20) ES_iAcquisition and DS_jChannel state information, channel gain value, and DS between_jChannel state information and channel gain values with the base station; step 30) calculating the current ES_iAnd DS_jIn collaboration, ES_iOverhead and DS of_jThe overhead of (c); step 40) calculating the current ES_iAnd DS_jIn case of no cooperation, ES_iOverhead and DS of_jThe overhead of (c); step 50), generating a migration overhead matrix; and step 60) obtaining a matching matrix and multi-terminal cooperation calculation migration cost by using a Hungarian algorithm.

Description

Cooperative computing migration method between energy-starved terminal and flow-starved terminal

Technical Field

The invention belongs to the field of communication, and particularly relates to a cooperative computing migration method between an energy-starved terminal and a flow-starved terminal.

Background

In recent years, novel intelligent internet of things applications (such as intelligent video monitoring systems driven by deep learning, precision agriculture, smart homes, automatic driving, industrial automatic control systems, smart medical treatment and the like) are receiving wide attention, and meanwhile, contradictions between the intelligent internet of things applications and intelligent mobile terminals with limited resources are more and more prominent. Mobile Cloud Computing (MCC) integrates traditional wireless cellular network and internet services, and a Computation migration (Computation migration) technology is used to upload Computation-intensive and delay-sensitive tasks to a Cloud for processing, which naturally becomes an effective solution. Specifically, the computing migration technique in the mobile cloud computing scheme can bring the following outstanding advantages: 1) the energy consumption of the terminal is reduced by transferring the complex computing task to the cloud; 2) the performance of the terminal for high-resource-demand application processing is expanded, and the data storage capacity of the terminal is improved; 3) and higher reliability and safety are provided for the data of the intelligent mobile terminal. As such, the mobile cloud computing industry has been rapidly developing in recent years. The method not only enables the mobile terminal to break through hardware limitation and obtain elastic performance expansion, but also reduces energy consumption through calculation and migration and prolongs the service time of the battery.

However, the intelligent mobile terminal needs to consider consumption of energy and traffic resources (data volume) for calculation migration, and especially needs to consider more important consideration when the wireless channel quality is poor. On the other hand, due to the limited battery capacity and the different traffic packages, the heterogeneous phenomenon of energy and traffic resources in the smart mobile terminal is ubiquitous. Most terminals or energy usage budgets are limited or traffic usage budgets are limited. For situations where the energy usage budget is limited, the battery power may quickly be depleted when the smart mobile terminal is continuously running a computationally intensive application. In addition, some terminals have very small battery capacities themselves. In the case of limited traffic usage budget, the traffic budget is easily exhausted when the smart mobile terminal is executing data intensive (e.g., high definition video) applications, and the traffic cost beyond the package portion is very expensive. And the computing migration cannot be continued regardless of whether the energy usage budget or the traffic usage budget is exhausted. Therefore, on the premise of not depleting battery power or generating extra traffic cost, it is very meaningful to research how to make a traffic-starved terminal (ES) and an energy-starved terminal (DS) effectively perform computation migration to expand the application range of mobile cloud computing, and it is also an important problem in the art.

Disclosure of Invention

The invention aims to overcome the defects of the existing computing migration technology, provides a cooperative computing migration method between an energy-starved terminal and a flow-starved terminal, fully utilizes the superior resources of the flow-starved terminal and the energy-starved terminal to make up the respective inferior resources by considering the common phenomenon that the energy resources and the flow resources of an intelligent mobile terminal are heterogeneous in practice, and effectively computes and migrates on the premise of not consuming the electric quantity of a battery or generating extra flow cost.

In order to solve the technical problems, the invention adopts the following technical scheme:

in a mobile cellular network based on D2D communication, a base station is located in the center of the network, a plurality of intelligent terminals are randomly distributed, the intelligent terminals migrate self calculation intensive tasks to the cloud for processing, the intelligent terminals comprise a flow shortage terminal ES and an energy shortage terminal DS, a flow shortage terminal set is represented as I {1, …, I, … and M }, wherein M is a positive integer, and I is any positive integer between [1 and M ]; the energy-starved terminal DS set is represented by J ═ {1, …, J, …, K }, where K is a positive integer, and J is any positive integer between [1, K ]; the method comprises the following steps:

step 10) the initiator passes the control messageBroadcasting demand signaling, when the helper responds to the initiator, collaboration starts, ES_iObtaining DS through control channel_jTask parameter of, DS_jObtaining ES through control channel_iThe task parameters of (1); the task parameters comprise data size, maximum allowable task delay and required CPU period; when the initiator is ES_iWhen, the helper is DS_j(ii) a When the initiator is DS_jThe helper is ES_i；ES_iIndicating the ith traffic starvation terminal, DS, in the system_jIndicating the jth energy starvation terminal in the system;

step 20) Using the channel estimation and feedback method, ES_iAcquisition and DS_jChannel state information h between_ijChannel gain value | h_ij|²And DS_jChannel state information h between base station and base station_jDAnd channel gain value | h_jD|²；

Step 30) calculating the current ES_iAnd DS_jIn collaboration, ES_iOverhead of

And DS_jOverhead of

The overhead comprises energy overhead and traffic overhead;

step 40) calculating the current ES_iAnd DS_jIn case of no cooperation, ES_iOverhead of

And DS_jOverhead of

Step 50) generating a migration overhead matrix C_M×K；

Step 60) the migration cost matrix C generated according to step 50)_M×KAnd acquiring a matching matrix and multi-terminal cooperation calculation migration overhead by using the Hungarian algorithm.

Preferably, in the step 30), ES_iOverhead in collaboration scenarios

As shown in formula (1):

wherein the content of the first and second substances,

represents ES_iAt the same time as the DS_jThe energy overhead in the case of collaboration,

represents ES_iThe weight of the energy overhead is given to,

represents ES_iAt the same time as the DS_jThe traffic overhead in the case of collaboration,

represents ES_iWeight of traffic cost; due to the terminal ES_iIn cooperation without direct communication with the base station, the terminal ES_iTraffic overhead of

Is equal to 0 and is equal to 0,

wherein the content of the first and second substances,

represents ES_iT represents ES_iMigration delay of the task;

DS_joverhead in collaboration scenarios

As shown in formula (2)：

Wherein the content of the first and second substances,

represents DS_jIn and ES_iThe energy overhead in the case of collaboration,

represents DS_jIn and ES_iThe traffic overhead in the case of collaboration,

represents DS_jThe weight of the energy overhead is given to,

represents DS_jWeight of traffic cost; due to the DS_jThe terminal will bear the ES_iAnd DS_jData overhead for all migration tasks, therefore

Wherein the content of the first and second substances,

represents ES_iThe size of the data of the migration task,

represents DS_jData size of migration task.

As a preferable example, the step 30) further includes: calculating an overhead optimal value during terminal cooperation, specifically comprising:

due to ES_iThe terminal will assume the energy consumption of all tasks in the cooperation, DS_jThe terminal will bear all data consumption in the cooperation, ES in the cooperation case_iIs optimized overhead of

As shown in formula (4), DS_jIs optimized overhead of

As shown in formula (5):

wherein the content of the first and second substances,

as shown in formula (6):

in the formula (6), N₀Representing white noise power; w represents a transmission bandwidth of the channel;

represents DS_jA maximum allowable migration delay of the migration task, an

Represents DS_jThe allowable latency of the migration task;

represents DS_jCPU cycles required for the migration task; d_cRepresenting cloud computing power; η represents an energy conversion efficiency factor;

to representIn the power-split RF energy harvesting technique, in ES_iAnd DS_jIn the case of collaboration, DS_jThe optimal power division factor of (1); h is_ijRepresents ES_iAnd DS_jChannel state information between; h is_jDRepresents DS_jChannel state information with the base station;

is shown in ES_iAnd DS_jIn the case of collaboration, DS_jThe optimal acquisition energy distribution factor of (1),

for distributing the collected energy and forwarding the ES_iEnergy and upload DS of tasks_jThe energy of the self task is shown as the formula (7):

ζ＝{2vηH-(ν+2ω)}²formula (8c)

Wherein H ═ H_jD|²And is and

ES_iand DS_jOptimal overhead in collaboration

As shown in formula (3):

wherein the content of the first and second substances,

is shown in ES_iAnd DS_jIn the case of collaboration, ES_iOf optimum transmission power, i.e. ES_iThe lowest transmission power of;

represents ES_iA maximum allowable migration delay of the migration task, an

Represents ES_iThe allowed latency of the migration task is,

represents ES_iCPU cycles required for the migration task, D_cRepresenting cloud computing power.

As a preferred example, in the step 40), when ES is used_iAnd DS_jWhen not cooperating, calculating the terminal ES according to the formula (10)_iOverhead of

When ES_iAnd DS_jWhen not cooperating, the terminal DS is calculated according to the formula (11)_jOverhead of

Wherein the content of the first and second substances,

indicating a traffic starved terminal ES in a non-cooperative situation_iEnergy consumption of (2);

as shown in formula (12 a);

indicating a traffic starved terminal ES in a non-cooperative situation_iThe flow rate of (a) is consumed,

indicating an energy starved terminal DS in the non-cooperative case_jEnergy consumption of (2);

as shown in formula (12 b);

indicating an energy starved terminal DS in the non-cooperative case_jThe flow rate of (a) is consumed,

wherein h is_iDRepresents ES_iChannel state information with the base station; n is a radical of₀Representing white noise power.

As a preferred example, the step 50) specifically includes: setting M flow-starved terminals ES and K energy-starved terminals DS in the system, and a migration overhead matrix C_M×KThe generation steps are as follows:

step 501) initializing matrix C_M×KLet matrix C_M×KWherein the elements are all 1;

step 502) j ═ q₁+1，q₁Representing the number of times step 505) returns to step 502), q₁Is 0;

step 503) i ═ q₂+1，q₂Representing the number of times step 505) returns to step 503), q₂Is 0;

step 504) judging whether a condition is met, wherein the condition is

And is

When the condition is satisfied, then ES is added_iAnd DS_jOverhead optima at collaboration

Assigned to matrix C_M×KElement c in (1)_ij(ii) a When the condition is not satisfied, then assigning infinity to the matrix C_M×KElement c in (1)_ij；

Step 505) of determining j>If K is true, if not, i is judged>Whether M is established or not; if yes, ending the circulation and generating a migration overhead matrix C_M×K；

The judgment i>Whether M is true or not, including: if i>If M is not satisfied, returning to the step 503); if i>If M is true, return to step 502), and q) in step 503)₂The value is 0.

Compared with the prior art, the method has the following beneficial effects: different from the traditional calculation migration method in which the intelligent mobile terminal completely consumes various resources of the intelligent mobile terminal to independently complete calculation migration, the method and the system fully consider the attention of the migration terminal to the deficient resources, enable each terminal to utilize the superior resources to make up for the inferior resources by promoting the cooperation between the flow deficient terminal and the energy deficient terminal, reduce the cost of the deficient resources of the terminal, increase the migratable frequency at lower cost, and simultaneously reduce the influence of a large amount of cost of the deficient resources on other normal work of the terminal.

Drawings

The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention.

FIG. 1 is a system architecture diagram of an embodiment of the present invention;

FIG. 2 is a flow chart of an embodiment of the present invention;

FIG. 3 shows an ES cell according to an embodiment of the present invention_iAnd DS_jCarrying out matched bipartite graphs;

fig. 4 is a diagram illustrating a comparison between a non-cooperative migration cost and a cooperative migration cost in the example of the present invention, and a relationship between the number of intelligent terminals and the weight of the scarce resource on the cooperative migration cost.

Detailed Description

The technical contents of the present invention will be described in detail below with reference to the accompanying drawings.

The method is suitable for the mobile cellular network based on D2D communication, wherein intelligent terminals in the cells all have calculation tasks needing calculation migration, and energy resources and flow resources of the terminals are heterogeneous. FIG. 1 is a diagram of a system for resource sharing-based collaborative computing migration according to the present invention. In a mobile cellular network based on D2D communication, a traffic starved terminal ES₁～ES₄Demand and energy starvation terminal DS₁～DS₄And forming a cooperation group according to a matching-cooperation mechanism to complete the calculation migration. NC refers to a type of intelligent mobile terminal with insufficient energy flow or sufficient energy flow. It is not under consideration in our collaboration system, as it has no collaboration qualification or willingness to collaborate. Specifically, we use cooperative groups (ES)₁,DS₃) For example, when a collaboration group is formed, the ES₁Transmitting tasks to be migrated to the DS in the form of radio frequency signals through the D2D communication link₃，DS₃Power-split RF energy harvesting (i.e., a technique in which RF signal power is split by a power-split factor for energy harvesting and the remaining power is used for signal processing) technique is used to extract RF energy from ES₁Energy is collected from the radio frequency signal. Then, DS₃ES forwarding using harvested energy₁And tasks that need to be migrated to the BS by itself, this process consumes traffic. And finally, the BS migrates the received task to the cloud for processing through the optical fiber link, and the received task is transmitted back to the corresponding terminal after the cloud processing is finished.

The system is applicable to a mobile cellular network based on D2D communication, wherein a base station BS is positioned in the center of the network, a plurality of intelligent terminals are randomly distributed, and the intelligent terminals migrate self calculation intensive tasks to the cloud for processing. The intelligent terminal comprises a flow starvation terminal ES and an energy starvation terminal DS. The set of low traffic terminals is denoted as I ═ {1, …, I, …, M }, where M is a positive integer and I is any positive integer between [1, M ]. The set of energy-starved terminals DS is denoted by J ═ {1, …, J, …, K }, where K is a positive integer and J is any positive integer between [1, K ]. The terminal with insufficient traffic is an intelligent terminal with insufficient monthly traffic budget and sufficient self battery power. The energy shortage terminal is an intelligent terminal with a shortage of battery power and a sufficient monthly traffic budget.

As shown in fig. 2, a method for collaborative computation migration between an energy-starved terminal and a traffic-starved terminal according to an embodiment of the present invention includes:

step 10) the initiator broadcasts the demand signaling through the control channel, when the helper responds to the initiator, the helper responds to the initiatorTo start, ES_iObtaining DS through control channel_jTask parameter of, DS_jObtaining ES through control channel_iThe task parameters of (1); the task parameters comprise data size, maximum allowable task delay and required CPU period; when the initiator is ES_iWhen, the helper is DS_j(ii) a When the initiator is DS_jThe helper is ES_i；ES_iIndicating the ith traffic starvation terminal, DS, in the system_jIndicating the jth energy starvation terminal in the system.

Step 20) Using the channel estimation and feedback method, ES_iAcquisition and DS_jChannel state information h between_ijChannel gain value | h_ij|²And DS_jChannel state information h with base station BS_jDAnd channel gain value | h_jD|²。

Step 30) calculating the current ES_iAnd DS_jIn collaboration, ES_iOverhead of

And DS_jOverhead of

The overhead includes energy overhead and traffic overhead. ES (ES)_iAnd DS_jWhen cooperating, the respective overhead may be expressed as a sum of the respective energy overhead and traffic overhead.

In step 30), ES_iOverhead in collaboration scenarios

As shown in formula (1):

wherein the content of the first and second substances,

represents ES_iAt the same time as the DS_jAbility to cooperateThe amount of overhead is measured,

represents ES_iThe weight of the energy overhead is given to,

represents ES_iAt the same time as the DS_jThe traffic overhead in the case of collaboration,

represents ES_iWeight of traffic cost; due to the terminal ES_iIn cooperation without direct communication with the base station, the terminal ES_iTraffic overhead of

Is equal to 0 and is equal to 0,

wherein the content of the first and second substances,

represents ES_iT represents ES_iMigration latency of tasks.

DS_jOverhead in collaboration scenarios

As shown in formula (2):

wherein the content of the first and second substances,

represents DS_jIn and ES_iThe energy overhead in the case of collaboration,

represents DS_jIn and ES_iThe traffic overhead in the case of collaboration,

represents DS_jThe weight of the energy overhead is given to,

represents DS_jWeight of traffic cost; due to the DS_jThe terminal will bear the ES_iAnd DS_jData overhead for all migration tasks, therefore

Wherein the content of the first and second substances,

represents ES_iThe size of the data of the migration task,

represents DS_jData size of migration task.

Preferably, considering that the multi-terminal cooperation calculation migration can be decomposed into a matching problem and a two-terminal cooperation problem to be performed respectively, the marketing can be split according to the optimal two-terminal cooperation scheme

And

further shown. Step 30) further comprises: calculating an overhead optimal value during terminal cooperation, specifically comprising:

due to ES_iThe terminal will assume the energy consumption of all tasks in the cooperation, DS_jThe terminal will bear all data consumption in the cooperation, ES in the cooperation case_iIs optimized overhead of

As shown in formula (4), DS_jIs optimized overhead of

As shown in formula (5):

wherein the content of the first and second substances,

as shown in formula (6):

in the formula (6), N₀Representing white noise power; w represents a transmission bandwidth of the channel;

represents DS_jA maximum allowable migration delay of the migration task, an

Represents DS_jThe allowable latency of the migration task;

represents DS_jCPU cycles required for the migration task; d_cRepresenting cloud computing power; η represents an energy conversion efficiency factor;

in the technique of representing power split RF energy harvesting, in ES_iAnd DS_jIn the case of collaboration, DS_jThe optimal power division factor of (1); h is_ijRepresents ES_iAnd DS_jChannel state information between; h is_jDRepresents DS_jChannel state information with the base station;

is shown in ES_iAnd DS_jIn the case of collaboration, DS_jThe optimal acquisition energy distribution factor of (1),

for distributing the collected energy and forwarding the ES_iEnergy and upload DS of tasks_jThe energy of the self task is shown as the formula (7):

ζ＝{2vηH-(ν+2ω)}²formula (8c)

Wherein H ═ H_jD|²And is and

ES_iand DS_jOptimal overhead in collaboration

As shown in formula (3):

wherein the content of the first and second substances,

is shown in ES_iAnd DS_jIn the case of collaboration, ES_iOf optimum transmission power, i.e. ES_iThe lowest transmission power of;

represents ES_iA maximum allowable migration delay of the migration task, an

Represents ES_iThe allowed latency of the migration task is,

represents ES_iCPU cycles required for the migration task, D_cRepresenting cloud computing power.

Step 40) calculating the current ES_iAnd DS_jIn case of no cooperation, ES_iOverhead of

And DS_jOverhead of

In step 40), when ES_iAnd DS_jWhen not cooperating, calculating the terminal ES according to the formula (10)_iOverhead of

When ES_iAnd DS_jWhen not cooperating, the terminal DS is calculated according to the formula (11)_jOverhead of

Wherein the content of the first and second substances,

indicating a traffic starved terminal ES in a non-cooperative situation_iEnergy consumption of (2);

as shown in formula (12 a);

indicating a traffic starved terminal ES in a non-cooperative situation_iThe flow rate of (a) is consumed,

indicating an energy starved terminal DS in the non-cooperative case_jEnergy consumption of (2);

as shown in formula (12 b);

indicating an energy starved terminal DS in the non-cooperative case_jThe flow rate of (a) is consumed,

wherein h is_iDRepresents ES_iChannel state information with the base station; n is a radical of₀Representing white noise power.

Step 50) generating a migration overhead matrix C_M×K. Step 50) specifically comprises: setting M flow-starved terminals ES and K energy-starved terminals DS in the system, and a migration overhead matrix C_M×KThe generation steps are as follows:

step 501) initializing matrix C_M×KLet matrix C_M×KWherein the elements are all 1;

step 502) j ═ q₁+1，q₁Representing the number of times step 505) returns to step 502), q₁Is 0;

step 503) i ═ q₂+1，q₂Representing the number of times step 505) returns to step 503), q₂Is 0;

step 504) judging whether a condition is met, wherein the condition is

And is

When the condition is satisfied, then ES is added_iAnd DS_jOverhead optima at collaboration

Assigned to matrix C_M×KElement c in (1)_ij(ii) a When the condition is not satisfied, then assigning infinity to the matrix C_M×KElement c in (1)_ij；

Step 505) of determining j>If K is true, if notImmediately, determine i>Whether M is established or not; if yes, ending the circulation and generating a migration overhead matrix C_M×K；

The judgment i>Whether M is true includes: if i>If M is not satisfied, returning to the step 503); if i>If M is true, return to step 502), and q) in step 503)₂The value is 0.

Step 60) the migration cost matrix C generated according to step 50)_M×KAnd acquiring a matching matrix and multi-terminal cooperation calculation migration overhead by using the Hungarian algorithm.

According to the matching mode shown in fig. 3, the Hungarian algorithm (matching is performed when the migration cost is the minimum) is used for solving the matching sub-problem, obtaining a matching matrix and calculating the migration cost through multi-terminal cooperation at the moment. Specifically, ES₁First of all with DS respectively₁、DS₂Up to DS_kMatching and calculating collaboration costs, preserving combinations (ES) with lowest costs₁，DS_j) Completing matching; then ES₂Same as except DS_jOther DS devices are matched and selected for ES in the same manner₂The optimal matching combination is combined, and the matching of all the terminals is finished by the same way.

The invention fully combines the resource sharing concept, considers the high concern of the migrating terminal on the deficient resources, ensures that each terminal can utilize the superior resources to make up the inferior resources by promoting the cooperation between the flow deficient terminal and the energy deficient terminal, reduces the expense of the deficient resources of the terminal, increases the migratable frequency with lower cost, and simultaneously reduces the influence of a large amount of expense of the deficient resources on other normal work of the terminal. In addition, the terminal only needs to allocate a channel by the AP after forming the cooperation (namely, each device needs to allocate a channel for uploading when not cooperating, and each cooperation group, namely two devices, after forming the cooperation allocates a channel, the occupied spectrum resource is reduced, thereby the overall performance of the network is also improved.

The embodiment of the invention relates to a cooperative computing migration technology based on resource sharing between terminals with energy shortage and flow shortage. Specifically, the embodiment of the invention provides a cooperative computing migration method in which an energy-starved terminal performs wireless energy collection from a received radio frequency signal of a traffic-starved terminal, and forwards a computing task (consumed traffic) of the energy-starved terminal to a cloud server as a return to achieve a win-win situation.

In the above embodiment, under the actual situation that the energy resource and the traffic resource of the intelligent mobile terminal are generally heterogeneous, if the two types of terminals (i.e., the energy-starved terminal and the traffic-starved terminal) are considered to form cooperation, the dominant resource of the terminal is used for replacing the dominant resource, and finally a win-win situation is formed between the two types of terminals, the two types of terminals can greatly increase the calculation and migration frequency at a low cost and cannot exhaust the deficient resource of the terminal to cause the terminal to fail to work normally. Specifically, the initiator (ES or DS) needs to broadcast a demand signaling through the control channel, and when a helper (DS or ES) responds to it, the cooperative process starts. Firstly, the ES sends a task to be migrated to the DS through D2D communication, and the DS performs energy collection from the radio frequency signals of the ES by using a power division radio frequency energy collection technology (namely, a technology of dividing the power of the radio frequency signals by a power division factor to perform energy collection, and using the rest power for signal processing). The DS then forwards the ES and its own tasks that need to be migrated to the Base Station (BS) using the collected energy, which consumes traffic. And finally, the base station transfers the received task to the cloud for processing. Through resource sharing, intelligent terminals forming a cooperation group do not need to consider overhead burden of deficient resources, and therefore the cooperation calculation migration method based on resource sharing between the terminals with energy deficiency and flow deficiency has research value.

A specific example will be provided below.

Consider a group of intelligent mobile terminals randomly distributed under the coverage of one base station BS. For ES_iAnd DS_jTo say, they all implement the document "Cloud-Vision: real-time face registration using a mobile-group access architecture [ C ]]IEEE, 2012: 000059-. | h_ij|²、|h_jD|²And | h_iD|²Obeying a mean value of λ_ij、λ_jDAnd λ_iDAre distributed exponentially, and the mean values are used separately

And

and (4) showing. Where ρ represents a path loss factor. The specific parameter settings in the calculations are given in table 1.

TABLE 1

Parameter(s)	Value of
		BS coverage	Radius R20 m
Number of intelligent terminals N	10～40
		Channel bandwidth W	5MHz
White noise power N0	-100dBm
		Path loss factor ρ	2.7
Energy conversion efficiency factor eta	0.7
		Weight mu1、φ1、μ2And phi2	0.6、0.4、0.7、0.3
Data size	0.42MB
		Maximum allowed time delay	600ms
Required CPU cycle	1000Megacycles
		Cloud computing power Dc	100GHz

In addition, a compute migration method under a non-collaborative scheme is provided for comparison. In the non-cooperative scheme, all intelligent mobile terminals migrate calculation tasks which need to be uploaded to a base station, and each terminal needs to consume energy and flow resources in the process. The parameters used in this protocol are the same as in the table above.

In the present collaborative computing migration example, the weight μ₁、φ₁、μ₂And phi₂Four different schemes are used, and the weight values for the first scheme are listed in table 1. In the second scheme, mu₁＝0.55、φ₁＝0.45、μ₂＝0.75、φ₂0.25. In the third scheme, mu₁＝0.5、φ₁＝0.5、μ₂＝0.8、φ₂0.2. In the fourth scheme, mu₁＝0.45、φ₁＝0.55、μ₂＝0.85、φ₂＝0.15。

The above-described cooperation scheme and the four schemes of the present cooperation calculation migration are subjected to multiple monte carlo simulations under the condition that the number of terminals N is 10, 15, …, and 40, respectively. All terminals are randomly distributed in the circular area, and experiments under different weights and different terminal numbers are repeated for 100 times respectively and averaged. The simulation results are shown in fig. 4. As can be seen from fig. 4, the total cost of computing migration under the collaborative scheme is always lower than that of the non-collaborative scheme. With starvation resource weight phi₁And mu₂The overhead of the collaborative computing migration is reduced. This is mainly due to the increased weight phi₁And mu₂Starved resources are given a higher value and thus saving starved resources will greatly reduce the cost of computation migration.

Based on the analysis, the method can effectively reduce the cost of the flow-starved terminal and the energy-starved terminal for the starved resources, and further reduce the overall calculation migration cost in the system.

Claims (2)

1. A cooperative computing migration method between an energy-starved terminal and a flow-starved terminal is characterized in that in a mobile cellular network based on D2D communication, a base station is located in the center of the network, a plurality of intelligent terminals are randomly distributed, the intelligent terminals migrate self computing-intensive tasks to the cloud for processing, each intelligent terminal comprises a flow-starved terminal ES and an energy-starved terminal DS, a flow-starved terminal set is represented as I {1, …, I, … and M }, wherein M is a positive integer, and I is any positive integer between [1 and M ]; the energy-starved terminal DS set is represented by J ═ {1, …, J, …, K }, where K is a positive integer, and J is any positive integer between [1, K ];

the method comprises the following steps:

step 10) the initiator broadcasts the demand signaling over the control channel, when the helper responds to the initiator, collaboration starts, ES_iObtaining DS through control channel_jTask parameter of, DS_jObtaining ES through control channel_iThe task parameters of (1); the task parameters comprise data size, maximum allowable task delay and required CPU period; when the initiator is ES_iWhen, the helper is DS_j(ii) a When the initiator is DS_jThe helper is ES_i；ES_iIndicating the ith traffic starvation terminal, DS, in the system_jIndicating the jth energy starvation terminal in the system;

step 20) Using the channel estimation and feedback method, ES_iAcquisition and DS_jChannel state information h between_ijChannel gain value | h_ij|²And DS_jChannel state information h between base station and base station_jDAnd channel gain value | h_jD|²；

Step 30) calculating the current ES_iAnd DS_jIn collaboration, ES_iOverhead of

And DS_jOverhead of

The overhead comprises energy overhead and traffic overhead; the step 30) further comprises: calculating an overhead optimal value during terminal cooperation, specifically comprising:

due to ES_iThe terminal will assume the energy consumption of all tasks in the cooperation, DS_jThe terminal will bear all data consumption in the cooperation, ES in the cooperation case_iIs optimized overhead of

As shown in formula (4), DS_jIs optimized overhead of

As shown in formula (5):

wherein the content of the first and second substances,

as shown in formula (6):

in the formula (6), N₀Representing white noise power; w represents a transmission bandwidth of the channel;

represents DS_jA maximum allowable migration delay of the migration task, an

Represents DS_jThe allowable latency of the migration task;

represents DS_jCPU cycles required for the migration task; d_cRepresenting cloud computing power; η represents an energy conversion efficiency factor;

representation power division radio frequency energy collection technologyIn ES_iAnd DS_jIn the case of collaboration, DS_jThe optimal power division factor of (1); h is_ijRepresents ES_iAnd DS_jChannel state information between; h is_jDRepresents DS_jChannel state information with the base station;

is shown in ES_iAnd DS_jIn the case of collaboration, DS_jThe optimal acquisition energy distribution factor of (1),

for distributing the collected energy and forwarding the ES_iEnergy and upload DS of tasks_jThe energy of the self task is shown as the formula (7):

ζ＝{2vηH-(n+2w)}²formula (8c)

Wherein H ═ H_jD|²And is and

ES_iand DS_jOptimal overhead in collaboration

As shown in formula (3):

wherein the content of the first and second substances,

is shown in ES_iAnd DS_jIn the case of collaboration, ES_iOf optimum transmission power, i.e. ES_iThe lowest transmission power of;

represents ES_iA maximum allowable migration delay of the migration task, an

Represents ES_iThe allowed latency of the migration task is,

represents ES_iCPU cycles required for the migration task, D_cRepresenting cloud computing power;

in said step 30), ES_iOverhead in collaboration scenarios

As shown in formula (1):

wherein the content of the first and second substances,

represents ES_iAt the same time as the DS_jThe energy overhead in the case of collaboration,

represents ES_iThe weight of the energy overhead is given to,

represents ES_iAt the same time as the DS_jThe traffic overhead in the case of collaboration,

represents ES_iWeight of traffic cost; due to the terminal ES_iIn cooperation without direct communication with the base station, the terminal ES_iTraffic overhead of

Is equal to 0 and is equal to 0,

wherein the content of the first and second substances,

represents ES_iT represents ES_iMigration delay of the task;

DS_joverhead in collaboration scenarios

As shown in formula (2):

wherein the content of the first and second substances,

represents DS_jIn and ES_iThe energy overhead in the case of collaboration,

represents DS_jIn and ES_iThe traffic overhead in the case of collaboration,

represents DS_jThe weight of the energy overhead is given to,

represents DS_jWeight of traffic cost; due to the DS_jThe terminal will bear the ES_iAnd DS_jData overhead for all migration tasks, therefore

Wherein the content of the first and second substances,

represents ES_iThe size of the data of the migration task,

represents DS_jData size of migration task;

step 40) calculating the current ES_iAnd DS_jIn case of no cooperation, ES_iOverhead of

And DS_jOverhead of

In the step 40), when ES_iAnd DS_jWhen not cooperating, the rootCalculating the terminal ES according to equation (10)_iOverhead of

When ES_iAnd DS_jWhen not cooperating, the terminal DS is calculated according to the formula (11)_jOverhead of

Wherein the content of the first and second substances,

indicating a traffic starved terminal ES in a non-cooperative situation_iEnergy consumption of (2);

as shown in formula (12 a);

indicating a traffic starved terminal ES in a non-cooperative situation_iThe flow rate of (a) is consumed,

indicating an energy starved terminal DS in the non-cooperative case_jEnergy consumption of (2);

as shown in formula (12 b);

indicating an energy starved terminal DS in the non-cooperative case_jThe flow rate of (a) is consumed,

wherein h is_iDRepresents ES_iChannel state information with the base station; n is a radical of₀Representing white noise power;

step 50) generating a migration overhead matrix C_M×K；

Step 60) the migration cost matrix C generated according to step 50)_M×KAnd acquiring a matching matrix and multi-terminal cooperation calculation migration overhead by using the Hungarian algorithm.

2. The method for collaborative computing migration between an energy-starved terminal and a traffic-starved terminal according to claim 1, wherein the step 50) specifically comprises: setting M flow-starved terminals ES and K energy-starved terminals DS in the system, and a migration overhead matrix C_M×KThe generation steps are as follows:

step 501) initializing matrix C_M×KLet matrix C_M×KWherein the elements are all 1;

step 502) j ═ q₁+1，q₁Representing the number of times step 505) returns to step 502), q₁Is 0;

step 503) i ═ q₂+1，q₂Representing the number of times step 505) returns to step 503), q₂Is 0;

step 504) judging whether a condition is met, wherein the condition is

And is

When the condition is satisfied, then ES is added_iAnd DS_jOverhead optima at collaboration

Assigned to matrix C_M×KElement c in (1)_ij(ii) a When the condition is not satisfied, then assigning infinity to the matrix C_M×KElement c in (1)_ij；

Step 505) of determining j>If K is true, if not, i is judged>Whether M is established or not; if yes, ending the circulation and generating a migration overhead matrix C_M×K；

The judgment i>Whether M is true includes: if i>If M is not satisfied, returning to the step 503); if i>If M is true, return to step 502), and q) in step 503)₂The value is 0.

Images