CN106534333A - Bidirectional selection computing unloading method based on MEC and MCC - Google Patents

Abstract

The invention discloses a bidirectional selection computing unloading method based on MEC and MCC. The method comprises steps of: sending to a network side a request to unload an operation program when it is detected that the computing amount of the operation program exceeds a threshold; receiving a unloading sequence allocated by the network side according to the request; computing local energy consumption and unloading energy consumption based on the received unloading sequence, and further computing each unloading energy consumption benefit and sending the same to the network side; and computing a local operation delay and further computing a difference between each unloading delay and local computing delay and sending the same to the network side; receiving a unloding decision matrix obtained by the network side based on the energy consumption benefit and the delay benefit, and returning a decision message; and performing unloading according to the unloading decision message. Therefore, the bidirectional selection computing unloading method based on MEC and MCC can realize the multi-criterion, multi-objective and bidirectional selection computing amount unloading of terminal linear weighting and network side analytical hierarchy process.

Description

A kind of two-way choice based on MEC and MCC calculates discharging method

Technical field

The present invention relates to field of mobile computing, particularly relates to a kind of two-way choice based on MEC and MCC and calculates unloading Method.

Background technology

In recent years, with the growing of user data, the extensive access of internet of things equipment and the variation of business Cause being skyrocketed through of data traffic in active wireless network, the drastically expansion of data scale.Meanwhile, the function of mobile terminal by Step is strengthened, its function has also been not limited solely to the communications field, but becomes people's mobile entertainment, office, reading, calculating Powerful carrier.Therefore, the various high complicated processes for depending on terminal that third party provides, beginning apply to mobile computing in a large number In.But, the computing capability of terminal is limited by volume, and current battery technology does not yet have breakthrough progress, and this gives Mobile computing field brings huge pressure.

In legacy network, in order to solve the problems, such as the intensive in mobile computing, service provider is aided in using cloud computing The data that process big data quantity, terminal cannot be processed or cannot be stored, that is, move cloud computing service (MCC, Mobile Cloud Computing).Service provider is deployed in special services in cloud, and terminal unit sends information to service, and service completes computing Result is sent back to into terminal afterwards, and necessary data is stored beyond the clouds.In order to service the user of diverse geographic location, in the Internet Multilayered structure in, cloud data center be located at core net.And core net apart from terminal use farther out, user message need through Some jumps can be reached, and this has just inevitably resulted in larger time delay, and larger time delay may affect user's Experience.Meanwhile, traditional cloud computing unloading thinking cannot solve time delay and jitter problem, and the physical distance of Cloud Server is often very Far, caused by network traffics generally existing skewness weighing apparatus, shake cannot also be avoided.

For the problem that cloud computing runs into, academia propose a kind of new technology movement edge calculations (MEC, Mobile Edge Computing), i.e., electricity is provided nearby using Radio Access Network RAN (Radio Access Network) Credit household's required service and high in the clouds computing function, create a carrier grade service for possessing high-performance, low latency and high bandwidth Environment, in acceleration network, every content, service and application is quick-downloading, allows consumer to enjoy continual high network quality body Test.Mobile edge calculations are in 2014 by ETSI (European Telecommunications Standards Institute) stand to lead to one of key technology of 5G, meeting 5G network low delays, the business of high bandwidth, high throughput More and more important role is play in requirement.It obtained including Nokia, Intel, Huawei, in multiple communication factories such as emerging The concern of family.Mobile edge calculations mainly include mobile edge calculations server and mobile two parts of mist terminal, and it has this Groundization, closely, low delay, location aware, the characteristics of network context information can be obtained, it is by the cloud computing ability of distal end Move user side to, there is provided more ubiquitous disposal ability.From the point of view of network side, mobile mist is calculated can be by content and position point From mode data processing and distribution are balanced to into network edge, reduce core net and cloud data center load；Edge calculations section Point can be realized pretreatment, intelligence shunting and directly process, reduce the transmission burden of network, and lifted to the data of upload The response speed of business.From the point of view of user side, the extra process based on the virtualized flexible resource configuration provides of network function Ability, while the scheme of more energy-conservation can be provided.

At present, the calculating unloading based on MEC is mainly set about in terms of two.One side is mainly in IoT (Internet of Things) field, using the more powerful equipment of computing capability as IoT equipment gateway, be sensor section Point provides the process of data and upload function, but this edge calculations node capacity is limited, and reliability is not ensured, ripple Dynamic property is big, also faces safety and privacy concerns, therefore can only be used as small range, the Internet of Things solution party of low amount of calculation Case.Second aspect is by mobile edge calculations server, there is provided similar in the server of cloud computing, but is deployed in access In net, to realize low delay, highly reliable communication plan.The method of research MEC unloadings is all from time delay and energy consumption two at present Aspect is started with, and is used as the decision method for unloading by maximizing the time delay-Energy Consumption Evaluation function of terminal.These implementations are past Toward starting with from network side, for a user, the time delay of their offers, consumption information are, by linear process, not account for The hobby of user.Additionally, existing research approach simply considers the resource-constrained situation of service end, the not load to service end Consideration is made with user fairness, this causes the income of current unloading scheme limited, while the service experience of user does not have Improve.

As MEC technologies propose that the time is not long, also deployment is not completed MEC related infrastructure completely, thus MCC and MEC unloading schemes typically isolate, and which results in the significant wastage of resource, and the advantage of amount of calculation unloading does not also embody completely Out.

The content of the invention

In view of this, it is an object of the invention to propose that a kind of two-way choice based on MEC and MCC calculates discharging method, The amount of calculation unloading of the multiple criteria of the linear weighted sum network side step analysis of terminal, multiple target, two-way choice can be realized.

The two-way choice provided based on MEC and MCC based on the above-mentioned purpose present invention calculates discharging method, including step：

When the amount of calculation for detecting operation program exceedes thresholding, the request of the unloading operation program is sent to network side；

Receive the sequence that be available for unload of the network side according to the request distribution；

According to the sequence for being available for unloading for receiving, local energy consumption and unloading energy consumption are calculated, and then calculate and every kind of be available for unloading Carry energy consumption income and be sent to network side；Meanwhile, local computing time delay is calculated, and then calculates every kind of time delay for being available for unloading and this Ground computation delay difference is simultaneously sent to network side；

Receive network side and matrix is determined according to the unloading that the energy consumption income and time delay income are obtained, the unloading of loopback is sentenced Certainly message；

Unloaded according to unloading decision message.

In some embodiments of the invention, it is available for unloading sequence to include cloud computing, mobile edge calculations server, free time Mobile intelligent terminal, is expressed as：d_p=(d_cc,d_ec,d_fc)d_cc,d_ec,d_fc∈{0,1}；

Wherein, d_pTo be available for unloading collection vector, d_ccRepresent whether Cloud Server set can provide unloading, d_ecRepresent mobile side Can edge calculation server provide unloading service, d_fcIntelligent terminal's set load sharing amount of calculation is indicated whether；These three amounts are all Binary amount, uses 0,1 expression, and 0 represents and can not unload, and 1 represents unloading.

In some embodiments of the invention, described local energy consumption is, produced by cpu process operation, to use equation below It is determined that：

Wherein, E_locRepresent the energy consumption needed for local computing, c_progThe cpu cycle number for calculating certain service needed is represented, f_locThe speed of local cpu is represented, unit is cycle/s, P_locRepresent the power consumption of CPU in the unit interval；

Then, according to the sequence for being available for unloading, the different energy consumption functions for unloading targets in the sequence of calculation, that is, unload energy respectively Consumption, represents：

Wherein, Q represents the covariance matrix for sending symbol；D_TkIt is the bit number sent required for customer service type k；η_T Presentation code coefficient, H represent channel matrix,It is the covariance matrix of interference plus noise；

The energy consumption difference of unloading and local computing is calculated, the energy consumption of three kinds of unloading targets is the same, is expressed as：

In some embodiments of the invention, calculate the time needed for local computing program：

Wherein, t_locRepresent the time delay of local computing；

Calculate the time required for three kinds of target unloadings：

Under cloud computing scene, the total time of unloading represents：

t_{cc_off}=t_T+t_R+t_w+t_process

t_{cc_off}The total time delay required for cloud computing unloading is represented, it is made up of four parts, first two section (t_T,t_R) point It is not the transmission delay of wireless transmission and wireless reception；t_wIt is the transmission delay of wireline side, B_bottleFor bottleneck bandwidth；t_queIt is Queuing delay is related to the state of network；t_propIt is propagation delay, it is related to physical distance；Hop represents arrival Cloud Server institute The jumping figure of needs；t_processTo process the time required for user program, f_ccIt is cpu cycle that cloud computing server can be provided Number；

Access mobile edge calculations server to calculate under scene, the total time of unloading represents：

t_{mc_off}=t_T+t_R+t_process

Wherein, f_mcIt is cpu cycle number that MEC servers can be provided；

Used as calculating under unloading target scene, the total time of unloading represents idle mobile intelligent terminal：

t_{fc_off}=2 (t_T+t_R)+t_process

Wherein, f_fcIt is cpu cycle number that other functional terminals can be provided；

Finally, unloading time delay and local computing time delay difference are calculated, as the time delay of three kinds of unloading targets is different, is prolonged The time difference needs to calculate respectively, i.e.,：

Wherein, Δ tⁱRepresent the delay inequality of terminal use i.

In some embodiments of the invention, also include：

Increase itself subjective and objective preference to two revenue functions, calculated energy consumption income, time delay income are carried out It is pre- to adjudicate, and it is then sent to network side.

In some embodiments of the invention, user terminal is in terms of energy consumption, in order to increase the life span of user terminal, It is generally acknowledged that when user terminal residual capacity of battery itself is relatively low, being more likely to unload away high energy consumption process, its mathematics is public Formula is described as follows：

Wherein, g_e ⁱFor the energy consumption yield parameter after revision；β_eFor the subjective preferences degree coefficient of user, its value is by user Terminal oneself sets, and preferred value is β_e∈(0,2)。For objective preference coefficient, by dump energy and the ratio of total electricity α_TeDetermine.

Affect according to weber phenanthrene Zener formula is introduced preference of the user terminal on postponing.Set according to different business simultaneously Different business coefficients are put, its concrete formula is described as follows：

Wherein, g_t ⁱFor the time delay yield parameter after revision；β_t ^kFor the preference coefficient of business, k represents different industry Business species, the real time business such as Streaming Media are higher to the coefficient requirements, and scientific algorithm class business requires relatively low to this, and preferred value is β_t ^k ∈(0,2)；sgn(Δt_i) for sign function.Three unloading targets will distinguish computation delay income g_t ⁱ=(g_{t_cc} ⁱ, g_{t_mc} ⁱ, g_{t_fc} ⁱ)。

Further, the two parameters that user terminal is obtained, can just can bear, and only final income is being for just for a user, It is just valuable.Therefore, determine that unloading target can be illustrated with following expression：

dⁱ=d_p ⁱ∩uⁱ(gⁱ _t+gⁱ _e), d ∈ (0,1) (12)

Wherein, dⁱAvailability judgements of the user i to three unloading targets is represented, 0 represents negative income or the unloading of no authority To this target, 1 represents that income is just and has permission unloading.uⁱX () is jump function.

In some embodiments of the invention, the network side that receives is obtained according to the energy consumption income and time delay income Unloading determine matrix include：

The energy consumption income sent according to each user terminal and time delay income, it is intended that feasible zone S；

Standard deviation load parameter is obtained according to the load information of current network, while according to user side unloading in the past Information acquisition fairness parameter；

According to energy consumption income, time delay income and feasible zone, and standard deviation load parameter and fairness parameter, set up many mesh Mark, multiple criteria Optimized model；

According to standard AHP algorithm, solution is carried out to Optimized model and obtains unloading decision matrix.

In some embodiments of the invention, current load information is calculated as：

Wherein, L_cc,L_mc,L_fcIt is cloud computing server group respectively, moves edge calculations server zone and idle intelligent movable The loading condition of terminal group, usesRepresent, W represents actual workload, and Cap represents max calculation ability；

In addition, according to the target information of user's unloading in the past, whole calculating being calculated using Jain justice indexes and being unloaded Fairness in journey, uses Q_T-1∈ (0,1] represent；Wherein, normalization justice variable is set to：Wherein 0 expression should It is no in cycle that the user terminal is unloaded；For user terminal i, his fairness valueCause This according to Jain fairness formula is：

Wherein, Q_TIt is the fair index in this unloading cycle；By EXSMOOTH, obtain comprehensive justice and refer to Number：

Q=α_qQ_T-1+(1-α_q)Q_T

Wherein, α_qFor smoothing factor.

In some embodiments of the invention, it is described to set up following multiple target, multiple criteria Optimized model：

s.t.

In some embodiments of the invention, according to unloading decision message, user carries out code and splits, then according to fractionation Code afterwards carries out unloading operation.

From the above it can be seen that the two-way choice based on MEC and MCC that the present invention is provided calculates discharging method, energy The equipment and calculating, transmittability of MCC and MEC are enough comprehensively utilized, and multiobject unloading selecting party are provided for mobile terminal Case, the process time that task not only can be reduced significantly but also the energy consumption that mobile device can be reduced to greatest extent.

Description of the drawings

Fig. 1 calculates the schematic flow sheet of discharging method user side for the two-way choice in embodiment based on MEC and MCC；

Fig. 2 calculates the frame diagram of discharging method for the two-way choice in embodiment based on MEC and MCC；

Fig. 3 calculates the schematic flow sheet of discharging method network side for the two-way choice in embodiment based on MEC and MCC；

Fig. 4 is the flow process that a two-way choice referred in embodiment based on MEC and MCC calculates discharging method user side Schematic diagram.

Specific embodiment

To make the object, technical solutions and advantages of the present invention become more apparent, below in conjunction with specific embodiment, and reference Accompanying drawing, the present invention is described in more detail.

It should be noted that the statement of all uses " first " and " second " is for differentiation two in the embodiment of the present invention The parameter of the entity or non-equal of individual same names non-equal, it is seen that the convenience of " first " " second " only for statement, should not The restriction to the embodiment of the present invention is interpreted as, subsequent embodiment is no longer illustrated one by one to this.

As embodiments of the invention, refering to shown in Fig. 1, being two-way choice in the embodiment of the present invention based on MEC and MCC Calculate the schematic flow sheet of discharging method user side.User side is directed to, the two-way choice based on MEC and MCC is calculated unloads Support method includes：

Step 101, when the amount of calculation for detecting operation program exceedes thresholding, sends the unloading operation program to network side Request.

Used as embodiment, detection needs program amount of calculation f of computing_proc, if it exceeds the thresholding f for pre-setting_thre, then The request of the unloading operation program is sent to network side.Further, when more than the thresholding f for pre-setting_threWhen, can sentence Disconnected CSI (Channel Side Information, Channel Side Information) channel information situation, if channel conditions allow user with big The request of the unloading operation program is sent to network side then in the speed rates of outage capacity.Further, to network side When sending the request of the unloading operation program, periodicity T can be opened and report self information, prepare unloading calculation procedure.

As a preferred embodiment, user terminal need according to dump energy relative to total electricity ratio judging The whether process of the amount of calculation of open detection operation program.Specifically, user terminal counts remaining electricity E_Tre, and calculate Relative to total electricity (E_Tall) ratio,When the ratio is more than default proportion threshold value, then inspection can be opened Survey the process of the amount of calculation of operation program.

Step 102, receives the sequence that be available for unload of the network side according to the request distribution.

In embodiment, the business information that user terminal is opened according to itself is sent to network side solicited message.Namely Say, network side according to the authentication protocol of the user terminal, can verify the power of oriented which the target unloading of the user terminal Limit.Further, network side is opened according to the user terminal service and own net ability, respond the cycle of the user terminal Property report information, and return to the user terminal one be available for unload sequence.

Preferably, it is available for unloading sequence include (as shown in Figure 2) cloud computing, mobile edge calculations server, free time Mobile intelligent terminal is available for unloading situation.The present invention is different from independent cloud computing of the prior art and mobile edge unloading Scheme, it is proposed that unloading scheme includes the cloud computing server (CC) provided by cloud computing enterprise and operator, has operator to exist The mobile edge calculations server (EC) disposed in access network and through safety certification, the idle mobile intelligent terminal for authorizing (FC).Message is reported in service and own net ability that network side is opened according to terminal, the periodicity for responding user, returns to use One, family coarseness is available for unloading sequence, that is, be expressed as：d_p=(d_cc,d_ec,d_fc)d_cc,d_ec,d_fc∈{0,1}。

Wherein, d_pTo be available for unloading collection vector, represent that terminal is available for the service unit species set for unloading.d_ccRepresent cloud clothes Whether business device set can provide unloading, d_ecRepresent that can mobile edge calculations server provide unloading service, d_fcIndicate whether Intelligent terminal's set can be with load sharing amount of calculation.These three amounts are all binary amounts, use 0,1 expression, and 0 represents and can not unload, and 1 expression can To unload.This step screening is carried out, can kick reduces the amount of calculation of following flow process except the set do not held qualification.

Step 103, according to the sequence for being available for unloading for receiving, calculates local energy consumption and unloading energy consumption, and then calculates every Plant and be available for unloading energy consumption income and being sent to network side；Meanwhile, calculate local computing time delay, and then calculate and every kind of be available for what is unloaded Time delay is with local computing delay inequality and is sent to network side.

In one embodiment, for the program of local computing, energy consumption is, produced by cpu process operation, to can use Equation below determines：

Wherein, E_locRepresent the energy consumption needed for local computing, c_progThe cpu cycle number for calculating certain service needed is represented, f_locThe speed of local cpu is represented, unit is cycle/s, P_locRepresent the power consumption of CPU in the unit interval.

Then, according to the sequence for being available for unloading, the different energy consumption functions for unloading targets in the sequence of calculation, that is, unload energy respectively Consumption.Its power consumption mainly determined by the transmission power for being sent to base station, receives power and compare transmit power to be negligible.Other Energy consumption does not occur in terminal, it is not required that consider from end side.Therefore, user's energy consumption during unloading can be expressed as below：

Wherein, Q represents the covariance matrix for sending symbol, in MIMO (Multiple-Input Multiple- Output, MIMO) in for according to channel situation optimize sending metrix so that obtain higher transmit power in good channel, from And overall rate is lifted, the mark of Q then shows the transmission power of terminal.D_TkIt is the bit number sent required for customer service type k, In uninstall process is calculated, the D of different unloading principles_TkThe content of expression is different, and what is had is probably that virtual machine configuration and user are defeated Enter data, some possibility need to transmit whole user program.η_TPresentation code coefficient because the present invention calculate use it is aromatic Calculation of capacity formula, but as chnnel coding limits the transfer rate for being unable to reach aromatic capacity in real system, therefore make The gap of actual coding speed and Theoretical Rate is represented with η ∈ (0,1).H represents channel matrix,It is the association of interference plus noise Variance matrix, represents the power of interference.

In addition, calculating the energy consumption difference of unloading and local computing, as the energy consumption of three kinds of unloading targets is the same, only need Calculating once unloads energy consumption.Wherein i represents that the energy consumption of terminal use i is poor.If Δ EⁱJust, to show to calculate unloading meeting to terminal Certain energy consumption income is brought, that is, reduces the energy consumption of terminal.

In another embodiment, calculate the time needed for local computing program：

Wherein, t_locRepresent the time delay of local computing.

Meanwhile, calculate the time required for unloading.As unloading for the present invention is loaded with three unloading targets, and three unloadings The physical distance of target, transmittability have difference, therefore the transmission delay of each unloading, are not quite similar.Three will be divided below Plant situation introduction.

Under cloud computing scene, the total time of unloading can be represented with below equation：

t_{cc_off}=t_T+t_R+t_w+t_process

t_{cc_off}The total time delay required for cloud computing unloading is represented, it is made up of four parts, first two section (t_T,t_R) point It is not the transmission delay of wireless transmission and wireless reception.Here, present invention assumes that in discharge time section, channel, interference are simultaneously Without significant change.As the coded system that transmitting-receiving is adopted is otherwise varied, therefore η values are otherwise varied；Additionally, the number for sending back and forth According to being different, D is used respectively_Tk,D_RkRepresent.t_wIt is the transmission delay of wireline side, B_bottleFor bottleneck bandwidth, t_queIt is to queue up Time delay is related to the state of network；t_propIt is propagation delay, it is related to physical distance；Hop is represented required for arrival Cloud Server Jumping figure, it will be assumed that round path and network state are constant in uninstall process.t_processTo process user program Required time, f_ccIt is cpu cycle number that cloud computing server can be provided, this value is typically much deeper than the place of terminal itself Reason ability.

Access mobile edge calculations server to calculate under scene, the total time of unloading can be represented with below equation：

t_{mc_off}=t_T+t_R+t_process

Wherein, t_T,t_RIt is similar to cloud computing, as MEC servers are typically deployed in RAN, therefore the biography of wire link Defeated time delay is very short, is negligible.f_mcIt is cpu cycle number that MEC servers can be provided, this value is taken less than cloud computing Business device, more than terminal itself CPU abilities.

Used as calculating under unloading target scene, the total time of unloading can use below equation table to idle mobile intelligent terminal Show：

t_{fc_off}=2 (t_T+t_R)+t_process

Wherein, t_T,t_RIt is similar to cloud computing, as terminal is the institute in same RAN as the transmission of unloading target To also need to plus the transmission delay to discharging terminal, it is considered that this transmission delay and terminal are close in itself.f_fcIt is The cpu cycle number that other functional terminals can be provided is close with user ability itself.Therefore, unloading between the terminals will not There is too many time income, the mainly income of energy consumption and the unloading under heavy duty network is supplemented.

Finally, unloading time delay and local computing time delay difference are calculated, as the time delay of three kinds of unloading targets is different, is prolonged The time difference needs to calculate respectively, i.e.,：

Wherein, Δ tⁱRepresent the delay inequality of terminal use i.If Δ tⁱJust, to show that calculating unloading can bring one to terminal Fixed time income, that is, reduce the time delay of terminal, and final result is that a 1*3 is vectorial.

In a preferred embodiment, energy consumption income and delay inequality (time delay income) are sent to network side in user terminal Before, pre- judgement can be carried out.That is, user terminal carries out pre- judgement to calculated energy consumption income, time delay income, And increase itself subjective and objective preference to two revenue functions.It is lower that energy consumption yield parameter shows as remaining electricity, is more inclined to In the unloading scheme for obtaining energy consumption income.And according to Psychophysics method and class of business, the revenue function of time delay is carried out Modification, has taken into full account user to calculating the impression of unloading.Energy consumption income of the present invention, two parameters of time delay income, this Two parameters can just can be born, and for user terminal, only two parameters are for just, just valuable to user terminal, by two The addition of parameter, carries out pre- judgement.

Further, user terminal is in terms of energy consumption, in order to increase the life span of user terminal, it is considered that in user When terminal residual capacity of battery itself is relatively low, it is more likely to unload away high energy consumption process, its mathematical formulae is described as follows：

Wherein, g_e ⁱFor the energy consumption yield parameter after revision；β_eFor the subjective preferences degree coefficient of user, its value is by user Terminal oneself sets, and preferred value is β_e∈(0,2)。For objective preference coefficient, by dump energy and the ratio of total electricity α_TeDetermine.

Further, user terminal can be carried out to time delay revenue function according to Psychophysics method and class of business Modification.Psychophysicss are to study heart thing relation and be allowed to a branches of psychology of quantification, be to physical stimulation and it draw Rise feel carry out the psychological field of quantification study.The present invention introduces user terminal to prolonging according to by weber phenanthrene Zener formula Slow preference affects.Simultaneously according to the different business coefficient of different business settings, its concrete formula is described as follows：

Wherein, g_t ⁱFor the time delay yield parameter after revision；β_t ^kFor the preference coefficient of business, k represents different industry Business species, the real time business such as Streaming Media are higher to the coefficient requirements, and scientific algorithm class business requires relatively low to this, and preferred value is β_t ^k ∈(0,2)；sgn(Δt_i) for sign function.Three unloading targets will distinguish computation delay income g_t ⁱ=(g_{t_cc} ⁱ, g_{t_mc} ⁱ, g_{t_fc} ⁱ).This formula shows that quantity of stimulus is increased by geometrical progression and sensation amount is then increased by arithmetical seriess, due to time delay not only Only it is the index for evaluating QoS, the even more actual index experienced of user, therefore introduces the evaluation methodology of psychophysicss here to fill Divide the impression for user being considered to calculating unloading.From the point of view of formula, as Δ t_iValue when reach to a certain degree, to user's Actual impression affects to be more and more lower.Therefore, can avoid all being assigned to most calculating task using this formula On MEC servers, the overall efficiency of network is improved.

Further, the two parameters that user terminal is obtained, can just can bear, for a user, only final income for just, It is just valuable to user, therefore terminal determines that unloading target can be illustrated with following expression：

dⁱ=d_p ⁱ∩uⁱ(gⁱ _t+gⁱ _e), d ∈ (0,1) (12)

Wherein, dⁱAvailability judgements of the user i to current three unloading targets is represented, 0 represents negative income or no authority This target is unloaded to, 1 expression income is just and has permission unloading.uⁱX () is jump function.

After the completion of calculating, user is by dⁱ, gⁱ _t, gⁱ _eIt is sent to the next step judgement that network side waits network side.

Step 104, receives network side and determines matrix, loopback according to the unloading that the energy consumption income and time delay income are obtained Unloading decision message.

In embodiment, two parameters that network side is sent according to each user terminal, it is intended that feasible zone.Meanwhile, network Side needs collection network side information by core network element and mobile edge calculations server (MEC), mainly including current network Load information and user terminal in the past unloading target information, and unloaded calculating whole calculating using Jain justice indexes Fairness in journey.Then, the data by more than unload problem to network side and are optimized modeling, then by step analysis Method, combination of qualitative and quantitative analysis can process the indeterminable NP-hard problems of Optimization Modeling.Matrix is determined according to unloading is obtained, Loopback decision message is to terminal use respectively.

In one preferably embodiment, the parameter that network side is sent according to each user terminal, it is intended that feasible zone S should Feasible zone S is the 0-1 binary matrixs of a 3*s, and s represents the number of users for needing unloading.

Network side needs collection network side information by core network element and MEC servers, mainly includes current network Load information L_sWith target information Q of user's unloading in the past_T-1.Wherein, shown in being calculated as follows of current load information：

Wherein, L_cc,L_mc,L_fcIt is cloud computing server group respectively, moves edge calculations server zone and idle intelligent movable The loading condition of terminal group, usesRepresent, W represents actual workload, and Cap represents max calculation ability.Due to the free time Mobile intelligent terminal is constantly individual one by one as unloading target, therefore calculates whole free time mobile intelligent terminal to simplify Service unloading group is normally treated as one.M, N, V represent cloud computing server, mobile edge calculations server and sky respectively The quantity of not busy mobile intelligent terminal.The average loaded in the case of representing three kinds.Connect when there is new unloading request After receipts, L will be changed_cc,L_mc,L_fcValue, so as to change L_s.Load balancing there is provided a kind of cheap effectively transparent method and expand Exhibition the network equipment and server bandwidth, increase handling capacity, Strengthens network data-handling capacity, improve network motility and can With property, and L_sLack of uniformity is represented, therefore its value should be the smaller the better.

According to the target information of user's unloading in the past, it is possible to use Jain justice indexes entirely calculate uninstall process to calculate Middle fairness, uses Q_T-1∈ (0,1] represent.In default situations, it is believed that MEC servers can provide best resource clothes Business, Cloud Server takes second place, and is finally other idle mobile intelligent terminals.Their normalization justice variables are set to by the present invention：It is no in wherein 0 expression cycle that the user terminal is unloaded.For user terminal i, his public affairs Levelling valueTherefore can be obtained according to Jain fairness formula：

Wherein, Q_TIt is the fair index in this unloading cycle.By EXSMOOTH, comprehensive justice can be obtained Index：

Q=α_qQ_T-1+(1-α_q)Q_T (15)

Wherein, α_qFor smoothing factor, preferred value is 0.5.

Therefore, the terminal parameter for being obtained according to above step and network side parameter, can set up following multiple target, multiple criteria Optimized model：

s.t.

Wherein, total energy consumption revenue function and total time delay revenue function are the energy consumption income letter of each user terminal respectively Number result sum and time delay revenue function result sum, i.e.,Q is fairness index value model Enclose and be (0,1], represent that fairness is better closer to 1.L_sIt is the standard deviation of loading condition, it should more little to load more balanced and more balanced, because Inverse of this present invention using standard deviation.In restrictive condition, first restrictive condition is the restriction to transmit power, P_TIt is transmitting General power；Second restrictive condition derives from terminal, it is desirable to for the income of terminal is positive；3rd restrictive condition is negative The restriction of load, the general assignment amount of unloading is no more than the load capacity for servicing；4th restrictive condition is the restriction of feasible zone, is One 0-1 matrix, shows that network side judgement may only be selected in S；5th restrictive condition is bandwidth, transmission when The total bandwidth that the bandwidth that time is used is provided no more than base station.From the point of view of to sum up, this problem is a NP-hard problem, but Stable solution within the limited time, the embodiment of the present invention is needed to take the relatively low AHP Algorithm for Solving of complexity, can be rapid Obtain a result, this to reduce unloading total time delay be also helpful to.

According to AHP algorithms, each user terminal final selection is unloaded objective matrix as destination layer target O for the present invention (choose)；By Q,G_t,G_eBasis for estimation of four indexs as rule layer, by the specifically chosen of each user terminal As scheme layered scheme.Secondly, we construct multilevel iudge matrix two-by-two, calculate the relative importance of element under single criterion. Next, according to Consistent Matrix method, the logical consistency of test and judge thinking, when Consistency Ratio is less than random index When, then approve current result.Finally, it is determined that per layer of all factors are for the sequencing weight process of general objective relative importance, Carry out total hierarchial sorting.This process is carried out from top successively to the bottom.For top, its level is single The result that the result of sequence namely always sorts.The AHP algorithms that the present invention takes are the AHP algorithms of standard, it is therefore an objective to quickly, surely The fixed result for obtaining suboptimum.

Finally, the result for being drawn according to AHP algorithms, obtains unloading and determines matrix S', and this is the matrix of a 3*s', each Row at most have an element to be 1, and remaining element is 0.And 1 that a line is exactly the target of unloading.Network side is according to decision matrix S', loopback decision message is to terminal use respectively.

Step 105, is unloaded according to unloading decision message.

In embodiment, according to unloading decision message, user carries out code and splits, then carried out according to the code after fractionation Unloading operation.Specifically, can be that the data that program is run are separated with operation logic, only the load shedding needed for operation To server end, server end runs the data of user, obtains the result needed for user by shared code logic.Need Illustrate, when offered load is heavier or during more user, might have user and cannot obtain unloading target, i.e. his unloading Vector is one 0 vectorial, and now no matter what the decision of terminal is, does not carry out unloading operation, terminal local processing routine.

In the another aspect of the embodiment of the present invention, refering to shown in Fig. 3, being based on the double of MEC and MCC in the embodiment of the present invention To the schematic flow sheet for selecting calculating discharging method network side.It is directed to network side, the two-way choice based on MEC and MCC Calculating discharging method includes：

Step 301, obtains the request of the unloading operation program that user side sends.

Step 302, according to sequence d for being available for unloading of the request distribution_p=(d_cc,d_ec,d_fc)d_cc,d_ec,d_fc∈{0, 1 }, and it is sent to user side.

The service opened according to terminal as embodiment, network side and own net ability, the periodicity for responding user are converged Report message, returns to one coarseness of user and is available for unloading sequence, that is, be expressed as：d_p=(d_cc,d_ec,d_fc)d_cc,d_ec,d_fc∈ {0,1}。

Wherein, d_pTo be available for unloading collection vector, represent that terminal is available for the service unit species set for unloading.d_ccRepresent cloud clothes Whether business device set can provide unloading, d_ecRepresent that can mobile edge calculations server provide unloading service, d_fcIndicate whether Intelligent terminal's set can be with load sharing amount of calculation.

Step 303, receives the energy consumption income and time delay income of user side transmission.

Step 304, the energy consumption income sent according to each user terminal and time delay income, it is intended that feasible zone S.

In embodiment, according to the parameter that each user terminal sends, it is intended that feasible zone S, feasible zone S are 3*s 0-1 binary matrixs, s represent need unloading number of users.

Step 305, obtains standard deviation load parameter according to the load information of current network, while according to the user side mistake Remove the information acquisition fairness parameter for unloading.Specific implementation process includes：

Being calculated as follows for current load information is shown：

Wherein, L_cc,L_mc,L_fcIt is cloud computing server group respectively, moves edge calculations server zone and idle intelligent movable The loading condition of terminal group, usesRepresent, W represents actual workload, and Cap represents max calculation ability.Due to sky Not busy mobile intelligent terminal is constantly individual one by one as unloading target, therefore calculates whole intelligent movable end free time to simplify End service unloading group is normally treated as one.M, N, V represent respectively cloud computing server, mobile edge calculations server and The quantity of idle mobile intelligent terminal.The average loaded in the case of representing three kinds.Connect when there is new unloading request After receipts, L will be changed_cc,L_mc,L_fcValue, so as to change L_s。

In addition, according to the target information of user's unloading in the past, it is possible to use Jain justice indexes are unloaded to calculate entirely to calculate Fairness during load, uses Q_T-1∈ (0,1] represent.In default situations, it is believed that MEC servers can provide best Resource service, Cloud Server are taken second place, and are finally other idle mobile intelligent terminals.Their normalization justice variables are set by the present invention For：It is no in wherein 0 expression cycle that the user terminal is unloaded.For user terminal i, he Fairness valueTherefore can be obtained according to Jain fairness formula：

Wherein, Q_TIt is the fair index in this unloading cycle.By EXSMOOTH, comprehensive justice can be obtained Index：

Q=α_qQ_T-1+(1-α_q)Q_T (15)

Wherein, α_qFor smoothing factor, preferred value is 0.5.

Step 306, according to energy consumption income, time delay income and feasible zone, and standard deviation load parameter and fairness parameter, build Vertical multiple target, multiple criteria Optimized model.

Wherein, following multiple target, multiple criteria Optimized model are set up：

s.t.

Step 307, according to standard AHP algorithm, carries out solution and obtains unloading decision matrix to Optimized model.

In embodiment, according to AHP algorithms, the present invention each user terminal final selection is unloaded objective matrix as Destination layer target O (choose)；By Q,G_t,G_eBasis for estimation of four indexs as rule layer, by each user terminal It is specifically chosen as scheme layered scheme.Secondly, we construct multilevel iudge matrix two-by-two, calculate the phase of element under single criterion To importance.

Then, according to Consistent Matrix method, the logical consistency of test and judge thinking, when Consistency Ratio is less than consistent at random During property index, then current result is approved.Finally, it is determined that per layer of all factors are for the sequencing weight of general objective relative importance Process, carries out total hierarchial sorting.This process is carried out from top successively to the bottom.For top, its layer The result that the result of secondary single sequence namely always sorts.The AHP algorithms that the present invention takes are the AHP algorithms of standard, it is therefore an objective to fast Speed, it is stably obtained the result of suboptimum.

Step 308, determines matrix according to unloading, respectively each terminal use of loopback decision message to user side.

Wherein, the result for being drawn according to AHP algorithms, obtains unloading and determines matrix S', and this is the matrix of a 3*s', each Row at most have an element to be 1, and remaining element is 0.And 1 that a line is exactly the target of unloading.Network side is according to decision matrix S', loopback decision message is to terminal use respectively.

Refering to as shown in figure 4, unloading for one two-way choice calculating referred in embodiment based on MEC and MCC of the present invention The schematic flow sheet of support method user side.User side is directed to, the two-way choice based on MEC and MCC calculates discharging method Including：

Step 401, when the user terminal dump energy is more than default proportion threshold value relative to the ratio of total electricity, Then can be with the process of the amount of calculation of open detection operation program.

Step 402, when program amount of calculation f of computing_procDuring more than the thresholding for pre-setting, open periodically to network side Send the request of the unloading operation program.

Step 403, receives sequence d that be available for unload of the network side according to the request distribution_p=(d_cc,d_ec,d_fc)d_cc, d_ec,d_fc∈{0,1}。

Step 404, according to the sequence for being available for unloading for receiving, calculates local energy consumption and unloading energy consumption, and then calculates every Plant and be available for unloading energy consumption income, execution step 406.

In embodiment, for the program of local computing, energy consumption is, produced by cpu process operation, to can use as follows Formula determines：

Wherein, E_locRepresent the energy consumption needed for local computing, c_progThe cpu cycle number for calculating certain service needed is represented, f_locThe speed of local cpu is represented, unit is cycle/s, P_locRepresent the power consumption of CPU in the unit interval.

Then, according to the sequence for being available for unloading, the different energy consumption functions for unloading targets in the sequence of calculation, that is, unload energy respectively Consumption.Therefore, user's energy consumption during unloading can be expressed as below：

Wherein, Q represents the covariance matrix for sending symbol；D_TkIt is the bit number sent required for customer service type k；η_T Presentation code coefficient；H represents channel matrix,It is the covariance matrix of interference plus noise, represents the power of interference.

In addition, calculating the energy consumption difference of unloading and local computing, as the energy consumption of three kinds of unloading targets is the same, only need Calculating once unloads energy consumption.Wherein i represents that the energy consumption of terminal use i is poor.If Δ EⁱJust, to show to calculate unloading meeting to terminal Certain energy consumption income is brought, that is, reduces the energy consumption of terminal.

Step 405, according to the sequence for being available for unloading for receiving, calculates local computing time delay, and then calculates every kind of being available for The time delay of unloading and local computing delay inequality, execution step 406.

As embodiment, the time needed for local computing program is calculated：

Wherein, t_locRepresent the time delay of local computing.

Meanwhile, calculate the time required for unloading.As unloading for the present invention is loaded with three unloading targets, and three unloadings The physical distance of target, transmittability have difference, therefore the transmission delay of each unloading, are not quite similar.Three will be divided below Plant situation introduction.

Under cloud computing scene, the total time of unloading can be represented with below equation：

t_{cc_off}=t_T+t_R+t_w+t_process

Wherein, t_{cc_off}Represent the total time delay required for cloud computing unloading；t_wIt is the transmission delay of wireline side；B_bottleFor Bottleneck bandwidth；t_queIt is queuing delay, it is related to the state of network；t_propIt is propagation delay, it is related to physical distance；Hop is represented Reach the jumping figure required for Cloud Server；t_processTo process the time required for user program, f_ccIt is cloud computing server institute The cpu cycle number that can be provided.

Access mobile edge calculations server to calculate under scene, the total time of unloading can be represented with below equation：

t_mc_o_ff=t_T+t_R+t_pro_cess

Wherein, t_T,t_RIt is similar to cloud computing；f_mcIt is cpu cycle number that MEC servers can be provided.

Used as calculating under unloading target scene, the total time of unloading can use below equation table to idle mobile intelligent terminal Show：

t_fc_o_ff=2 (t_T+t_R)+t_pro_cess

Wherein, t_T, t_RIt is similar to cloud computing；f_fcIt is cpu cycle number that other functional terminals can be provided.

Finally, unloading time delay and local computing time delay difference are calculated, as the time delay of three kinds of unloading targets is different, is prolonged The time difference needs to calculate respectively, i.e.,：

Step 406, increases itself subjective and objective preference to two revenue functions, to calculated energy consumption income, time delay Income carries out pre- judgement, and is then sent to network side.Specific implementation process includes：

User terminal in terms of energy consumption, in order to increase the life span of user terminal, it is considered that in user terminal itself When residual capacity of battery is relatively low, it is more likely to unload away high energy consumption process, its mathematical formulae is described as follows：

Wherein, g_e ⁱFor the energy consumption yield parameter after revision；β_eFor the subjective preferences degree coefficient of user, its value is by user Terminal oneself sets, and preferred value is β_e∈(0,2)。For objective preference coefficient, by dump energy and the ratio of total electricity α_TeDetermine.

Affect according to weber phenanthrene Zener formula is introduced preference of the user terminal on postponing.Set according to different business simultaneously Different business coefficients are put, its concrete formula is described as follows：

Wherein, g_t ⁱFor the time delay yield parameter after revision；β_t ^kFor the preference coefficient of business, k represents different industry Business species, the real time business such as Streaming Media are higher to the coefficient requirements, and scientific algorithm class business requires relatively low to this, and preferred value is β_t ^k ∈(0,2)；sgn(Δt_i) for sign function.Three unloading targets will distinguish computation delay income g_t ⁱ=(g_{t_cc} ⁱ, g_{t_mc} ⁱ, g_{t_fc} ⁱ)。

Further, the two parameters that user terminal is obtained, can just can bear, and only final income is being for just for a user, It is just valuable.Therefore, determine that unloading target can be illustrated with following expression：

dⁱ=d_p ⁱ∩uⁱ(gⁱ _t+gⁱ _e), d ∈ (0,1) (12)

Wherein, dⁱAvailability judgements of the user i to three unloading targets is represented, 0 represents negative income or the unloading of no authority To this target, 1 represents that income is just and has permission unloading.uⁱX () is jump function.

After the completion of calculating, user is by dⁱ, gⁱ _t, gⁱ _eIt is sent to network side.

Step 407, receives network side and determines square according to the unloading that the energy consumption income and time delay income after pre- judgement is obtained Battle array, the unloading decision message of loopback.

Step 408, according to unloading decision message, user carries out code and splits, then unloaded according to the code after fractionation Carry operation.

In sum, a kind of two-way choice based on MEC and MCC that the present invention is provided calculates discharging method, creatively The bidirectional selection scheme based on MCC and MEC is employed, by the ability for introducing cloud computing and mobile edge calculations simultaneously, is used How far the unloading target at family is changed；Also, the infrastructure and computing resource of network are taken full advantage of, not only can be reduced significantly and be appointed The process time of business, can also reduce the energy consumption of mobile device to greatest extent；And, enhance the calculating unloading energy of network Power；At the same time, the different business according to user's application, provides extra preference function for terminal so that body in unloading decision Reveal the preference of the unloading of terminal use, improve the service experience of user；In addition, according to the standing state of network so that net The overall performance of network is improved, and increased the stability of network；Finally, the whole described two-way choice based on MEC and MCC Calculate discharging method and be compact, easy to control, with extensive, great dissemination.

Those of ordinary skill in the art should be understood：The discussion of any of the above embodiment is exemplary only, not It is intended to imply that the scope of the present disclosure (including claim) is limited to these examples；Under the thinking of the present invention, above example Or can also be combined between the technical characteristic in different embodiments, step can be realized with random order, and is existed such as Many other changes of the different aspect of the upper described present invention, for simple and clear their no offers in details.

Embodiments of the invention be intended to fall within the broad range of claims it is all such replace, Modification and modification.Therefore, all any omissions within the spirit and principles in the present invention, made, modification, equivalent, improvement Deng should be included within the scope of the present invention.

Claims (10)

1. a kind of two-way choice based on MEC and MCC calculates discharging method, it is characterised in that including step：

When the amount of calculation for detecting operation program exceedes thresholding, the request of the unloading operation program is sent to network side；

Receive the sequence that be available for unload of the network side according to the request distribution；

According to the sequence for being available for unloading for receiving, local energy consumption and unloading energy consumption are calculated, and then calculate and every kind of be available for unloading energy Consumption income is simultaneously sent to network side；Meanwhile, local computing time delay is calculated, and then every kind of time delay for being available for unloading is calculated with local meter Calculate delay inequality and be sent to network side；

Receive network side and matrix is determined according to the unloading that the energy consumption income and time delay income are obtained, the unloading judgement of loopback disappears Breath；

Unloaded according to unloading decision message.

2. method according to claim 1, it is characterised in that being available for unloading sequence includes cloud computing, mobile edge calculations Server, idle mobile intelligent terminal, are expressed as：d_p=(d_cc,d_ec,d_fc)d_cc,d_ec,d_fc∈{0,1}；

Wherein, d_pTo be available for unloading collection vector, d_ccRepresent whether Cloud Server set can provide unloading, d_ecRepresent mobile edge meter Calculate server and unloading service, d can be provided_fcIntelligent terminal's set load sharing amount of calculation is indicated whether；These three amounts are all binary Amount, uses 0,1 expression, and 0 represents and can not unload, and 1 represents unloading.

3. method according to claim 2, it is characterised in that described local energy consumption is produced by cpu process operation, Determined with equation below：

$E_{loc}=\frac{c_{prog}}{f_{loc}}\×P_{loc}$

Wherein, E_locRepresent the energy consumption needed for local computing, c_progRepresent the cpu cycle number for calculating certain service needed, f_locRepresent The speed of local cpu, unit are cycle/s, P_locRepresent the power consumption of CPU in the unit interval；

Then, according to the sequence for being available for unloading, the different energy consumption functions for unloading targets in the sequence of calculation, that is, unload energy consumption respectively, Represent：

$E_{off}=\frac{tr\left(Q\right)D_{Tk}}{{\mathrm{\η}}_T{\log }_2\det (I+{\mathrm{HQH}}^H{R}_n^{-1})}$

Wherein, Q represents the covariance matrix for sending symbol；D_TkIt is the bit number sent required for customer service type k；η_TRepresent Code coefficient, H represent channel matrix,It is the covariance matrix of interference plus noise；

The energy consumption difference of unloading and local computing is calculated, the energy consumption of three kinds of unloading targets is the same, is expressed as：

4. method according to claim 3, it is characterised in that calculate the time needed for local computing program：

$t_{loc}=\frac{c_{prog}}{f_{loc}}$

Wherein, t_locRepresent the time delay of local computing；

Calculate the time required for three kinds of target unloadings：

Under cloud computing scene, the total time of unloading represents：

t_{cc_off}=t_T+t_R+t_w+t_process

$t_T=\frac{D_{Tk}}{{\mathrm{\η}}_T{\log }_2\det (I+{\mathrm{HQH}}^H{R}_n^{-1})}$

$t_R=\frac{D_{Rk}}{{\mathrm{\η}}_R{\log }_2\det (I+{\mathrm{HQH}}^H{R}_n^{-1})}$

$t_w=(\frac{D_{Tk}}{B_{bottle}}+\frac{D_{Rk}}{B_{bottle}}+t_{que}+t_{prop})\×hop$

$t_{process}=\frac{c_{prog}}{f_{cc}}$

t_{cc_off}The total time delay required for cloud computing unloading is represented, it is made up of four parts, first two section (t_T,t_R) be respectively Wireless transmission and the transmission delay of wireless reception；t_wIt is the transmission delay of wireline side, B_bottleFor bottleneck bandwidth；t_queIt is to queue up Time delay is related to the state of network；t_propIt is propagation delay, it is related to physical distance；Hop is represented required for arrival Cloud Server Jumping figure；t_processTo process the time required for user program, f_ccIt is cpu cycle number that cloud computing server can be provided；

Access mobile edge calculations server to calculate under scene, the total time of unloading represents：

t_{mc_off}=t_T+t_R+t_process

$t_{process}=\frac{c_{prog}}{f_{mc}}$

Wherein, f_mcIt is cpu cycle number that MEC servers can be provided；

Used as calculating under unloading target scene, the total time of unloading represents idle mobile intelligent terminal：

t_{fc_off}=2 (t_T+t_R)+t_process

$t_{process}=\frac{c_{prog}}{f_{fc}}$

Wherein, f_fcIt is cpu cycle number that other functional terminals can be provided；

Finally, unloading time delay and local computing time delay difference are calculated, due to the time delays of three kinds of unloading targets it is different, delay inequality Need to calculate respectively, i.e.,：

${\mathrm{\Δt}}^i=(t_{loc}^i-t_{cc\_off}^i,t_{loc}^i-t_{mc\_off}^i,t_{loc}^i-t_{fc\_off}^i)$

Wherein, Δ tⁱRepresent the delay inequality of terminal use i.

5. method according to claim 1, it is characterised in that also include：

Increase itself subjective and objective preference to two revenue functions, anticipation is carried out to calculated energy consumption income, time delay income Certainly, and it is then sent to network side.

6. method according to claim 5, it is characterised in that characterized in that, user terminal is in terms of energy consumption, in order to increase Plus the life span of user terminal, it is considered that when user terminal residual capacity of battery itself is relatively low, it is more likely to high energy consumption Process is unloaded away, and its mathematical formulae is described as follows：

${g_e}^i={\mathrm{\β}}_e{\left(2e\right)}^{{\mathrm{\α}}_{iTe}-1}\×{\mathrm{\ΔE}}_i---\left(10\right)$

Wherein, g_e ⁱFor the energy consumption yield parameter after revision；β_eFor the subjective preferences degree coefficient of user, its value is by user terminal Oneself setting, preferred value is β_e∈(0,2)。For objective preference coefficient, by dump energy and the ratio cc of total electricity_TeCertainly It is fixed.

Affect according to weber phenanthrene Zener formula is introduced preference of the user terminal on postponing.Simultaneously business setting according to different not Same business coefficient, its concrete formula are described as follows：

${g_t}^i={{\mathrm{\β}}_t}^k\left({\log }_{2}e\right)ln(\frac{\left|{\mathrm{\Δt}}_i\right|}{t_{loc}}+1)\mathrm{sgn}\left({\mathrm{\Δt}}_i\right)---\left(11\right)$

Wherein, g_t ⁱFor the time delay yield parameter after revision；β_t ^kFor the preference coefficient of business, k represents different business kinds The real time business such as class, Streaming Media are higher to the coefficient requirements, and scientific algorithm class business requires relatively low to this, and preferred value is β_t ^k∈ (0,2)；sgn(Δt_i) for sign function.Three unloading targets will distinguish computation delay income g_t ⁱ=(g_{t_cc} ⁱ, g_{t_mc} ⁱ, g_{t_fc} ⁱ)。

Further, the two parameters that user terminal is obtained, can just can bear, and only final income is just, just have for a user Value.Therefore, determine that unloading target can be illustrated with following expression：

dⁱ=d_p ⁱ∩uⁱ(gⁱ _t+gⁱ _e), d ∈ (0,1) (12)

Wherein, dⁱAvailability judgements of the user i to three unloading targets is represented, 0 represents that negative income or no authority are unloaded to this mesh Mark, 1 expression income are just and have permission unloading.uⁱX () is jump function.

7. method according to claim 6, it is characterised in that described to receive network side according to the energy consumption income and prolong When income obtain unloading determine matrix include：

The energy consumption income sent according to each user terminal and time delay income, it is intended that feasible zone S；

Standard deviation load parameter is obtained according to the load information of current network, while according to the information of user side unloading in the past Obtain fairness parameter；

According to energy consumption income, time delay income and feasible zone, and standard deviation load parameter and fairness parameter, multiple target, many is set up Criterion Optimized model；

According to standard AHP algorithm, solution is carried out to Optimized model and obtains unloading decision matrix.

8. method according to claim 7, it is characterised in that current load information is calculated as：

$L_s=\sqrt{\underset{m=1}{\overset{M}{\Σ}}{({L_{cc}}^m-{\stackrel{\‾}{L}}_{cc})}^2+\underset{n=1}{\overset{N}{\Σ}}{({L_{mc}}^n-{\stackrel{\‾}{L}}_{mc})}^2+{(\frac{\underset{v=1}{\overset{V}{\Σ}}{L_{fc}}^v}{V}-{\stackrel{\‾}{L}}_{fc})}^2}$

Wherein, L_cc,L_mc,L_fcIt is cloud computing server group respectively, moves edge calculations server zone and idle intelligent movable end The loading condition of end group, usesRepresent, W represents actual workload, and Cap represents max calculation ability；

In addition, according to the target information of user's unloading in the past, whole calculating in uninstall process is calculated using Jain justice indexes Fairness, uses Q_T-1∈ (0,1] represent；Wherein, normalization justice variable is set to：Wherein 0 represents the cycle It is interior the user terminal not to be unloaded；For user terminal i, his fairness valueCause This according to Jain fairness formula is：

$Q_T\left(S\right)=\frac{{\left(\underset{i=1}{\overset{S}{\Σ}}{f}_i\right)}^2}{S\underset{i=1}{\overset{S}{\Σ}}{f_i}^2}$

Wherein, Q_TIt is the fair index in this unloading cycle；By EXSMOOTH, comprehensive justice index is obtained：

Q=α_qQ_T-1+(1-α_q)Q_T

Wherein, α_qFor smoothing factor.

9. method according to claim 7, it is characterised in that the following multiple target of the foundation, multiple criteria Optimized model：

$\begin{array}{cc}\underset{S_{3*S}}{max}& Q,\frac{1}{L_s},G_t,G_e\end{array}$

s.t.

$\left.\begin{array}{c}tr\left(Q\right)\≤P_T,Q\≥0\\ {g^i}_t+{g^i}_e>0\\ 0\≤L_{cc},L_{mc},L_{fc}\≤1\\ s_{ij}\∈S,s_{ij}\∈\{0,1\}\\ \underset{i=1}{\overset{S}{\Σ}}{B}_i\≤B\end{array}\right\}\mathrm{\χ}$

10. the method according to claim 1 to 9 any one, it is characterised in that according to unloading decision message, Yong Hushi Line code splits, and then carries out unloading operation according to the code after fractionation.