阅读说明：本技术 5g/b5g环境下移动社交云资源分配技术 (Mobile social cloud resource allocation technology in 5G/B5G environment ) 是由 王兴伟 代晨曦 任庆庆 易波 于 2021-06-21 设计创作，主要内容包括：本发明公开了一种5G/B5G环境下移动社交云资源分配技术,基于分解和支配的多目标优化资源分配方法,该方法构建了在5G/B5G环境中,面向社区的计算迁移场景下的移动社交云资源分配机制整体框架,并对系统各个模块的功能进行了介绍。此外,对系统的时延、能耗、社交虚拟货币支出和社交效益等多目标问题进行建模,设计基于分解和支配的多目标优化资源分配算法,本发明方法能够通过对系统多目标进行优化,从而有效提高了5G/B5G环境下云资源分配利用率和系统性能。(The invention discloses a mobile social cloud resource allocation technology in a 5G/B5G environment, and discloses a multi-objective optimization resource allocation method based on decomposition and domination, wherein an overall framework of a mobile social cloud resource allocation mechanism in a community-oriented computing migration scene in the 5G/B5G environment is constructed, and functions of all modules of a system are introduced. In addition, multi-objective problems of time delay, energy consumption, social virtual currency expenditure, social benefits and the like of the system are modeled, and a multi-objective optimization resource allocation algorithm based on decomposition and domination is designed.)

1. A mobile social cloud resource allocation technology in a 5G/B5G environment is characterized in that the resource allocation demand of a mobile social cloud in a computing migration scene in a 5G/B5G environment can be met according to resource transaction prices and a reasonable and effective resource allocation mechanism of time delay, energy consumption and social relations, the average time delay, the average energy consumption and the average expenditure of users are minimized by searching an optimal resource matching scheme, and the average social benefit of the users is maximized; the method mainly comprises the following steps:

step 1, determining a problem research scene, and carrying out abstract modeling on a mobile social cloud system facing a community;

step 2, determining an optimization target for optimizing the resource allocation method;

step 3, modeling the resource allocation problem in the mobile social cloud environment facing the community from four dimensions;

step 4, constructing sub-objective functions through four dimensions, and establishing a problem model;

and 5, based on the decomposed and dominated relation construction, optimizing the multi-objective cloud resource allocation method.

2. The mobile social cloud resource allocation technique under the 5G/B5G environment of claim 1, wherein: in the problem research scenario in step 1, the calculation migration in the community is faced, and after the mobile social cloud system facing the community is abstracted and simplified, the problem research scenario can be divided into three layers: resource domain, physical domain, and social domain.

3. The mobile social cloud resource allocation technique under the 5G/B5G environment of claim 1, wherein: the optimization goal in the step 2 is mainly considered from four dimensions, and the average time delay of the system, the average energy consumption and the average user expense are minimized and the average social benefit of the user is maximized by finding an optimal resource matching scheme.

4. The mobile social cloud resource allocation technique under the 5G/B5G environment of claim 1, wherein: the step 3 comprises the following steps:

step 3.1, when a demand side carries out calculation migration, the time loss of an intermediate link is inevitable, and the considered range of the system time delay comprises transmission time delay, queue waiting time delay, agent execution time delay, intermediate communication and result return time delay; because the calculation needing migration is usually that the self calculation amount is relatively large, and the result data amount is relatively small, the result return delay can be ignored, and the total delay characteristic vector is constructed as shown in the following formula (1):

wherein the content of the first and second substances,for the transmission delay of user i at time t,waiting and executing time delay for the user i at the time t;

step 3.2, the energy consumption of the calculation migration is concentrated in the calculation transmission process and the agent execution process; because the social cloud uses an asymmetric transmission mode, and in the asymmetric transmission mode, the uplink transmission rate is much lower than the downlink transmission rate, the energy consumption evaluation is divided into two blocks, and the constructed energy consumption vector is as shown in formula (2):

wherein the content of the first and second substances,represents the transmission energy consumption, and is defined as shown in formula (3):

P_iis the transmission power, furthermoreRepresenting the agent execution energy consumption, considering two user-level modes of migration, i.e. single agent migration and multi-agent migration, where the energy consumption is divided into two cases, single agent mode such as formula (4) and multi-agent mode such as formula (5), accordingly.

Wherein the content of the first and second substances,andrespectively representing the code scale of the idx-th calculation when the agent j executes 1bit code consumption, the code scale of the idx-th calculation when the demander i migrates to the agent j and the energy value consumed when the agent j runs 1CPU cycle;

step 3.3 virtual social Currency Cost model is intended to evaluate social virtual Currency costs according to user transaction amounts, and construct a virtual currency Cost spent by a requester i in a transaction at time t_i,tAs shown in formula (6):

wherein the content of the first and second substances,is cost price, Com_i,idx,j,tThe calculation amount of the idx calculation in the self to-be-migrated calculation that the demand party i migrates to the agent j at the time t is represented, and the unit is CPU cycle;

step 3.4, social benefit is social harvest brought to a demand party by a certain agent selection scheme, a measure of social relationship in the MSN is fully exerted, and a social benefit vector is constructed as shown in a formula (7):

wherein, b_iIs a weight coefficient, SR is social reputation, IntimD is intimacy, representing the strength of the social relationship between the social user and the agent, with a value range of [0,1]。

5. The mobile social cloud resource allocation technique under the 5G/B5G environment of claim 1, wherein: the specific steps of the step 4 comprise:

step 4.1, according to the task priority queue shown in the formula (8), determining that the system sub-target is average time delay minimization, and the objective function is shown in the formula (9):

step 4.2, determining the average energy consumption minimization objective function of the system sub-target users as formula (10):

step 4.3, determining the system sub-target user average virtual currency cost minimization objective function as (11):

step 4.4, determining the social average benefit maximization objective function of all users of the system sub-targets as formula (12):

in order to maintain consistency in the target optimization direction, the reciprocal of the above formula is converted into the formula (13):

step 4.5, by searching an optimal resource matching scheme, minimizing the average time delay of the system, the average energy consumption and the average user expense and maximizing the average social benefit of the user, the problem model target f (x) can be expressed as formula (14), formula (15) and formula (16):

minF(X)＝(f₁(X),...,f_m(X))^T (14)

X＝(x₁,...x_N)^T (15)

m＝4 (16)

wherein f is₁Representing the mean value of energy consumption of the user, f₂Representing the user's virtual social currency payout mean, f₃Representing the mean value of the user delay, f₄And the reciprocal of the social benefit mean value of the user is represented, X is a scheme set for resource allocation in a calculation migration scene, and m is the target number.

6. The mobile social cloud resource allocation technique under the 5G/B5G environment of claim 1, wherein: the specific steps of the step 5 are as follows:

step 5.1, converting the multi-target problems of time delay, energy consumption, social benefits and virtual social expenditure in each resource allocation scheme into a single-target problem which is easier to solve and understand through decomposition;

step 5.2, scheme sets are sorted quickly and non-dominantly, and Pareto domination hierarchical relations of the whole feasible scheme system are determined according to multiple targets such as time delay, energy consumption, social benefits and virtual social expenditure;

step 5.3, the non-dominated sorting aims at the feasible solutions in the solution set, namely feasible solutions, and introduces constraint violation values for the infeasible solutions to process the problem of the infeasible solutions.

Technical Field

The invention belongs to the technical research field of mobile social cloud resource allocation under the environment of 5G/B5G, and particularly relates to analysis of factors such as time delay and energy consumption of computing migration in mobile social cloud resource allocation and social factors in a mobile social network, and a multi-target-oriented mobile social cloud resource allocation mechanism.

Background

The invention provides a multi-objective optimization resource allocation method based on decomposition and domination, and aims at solving the problem that the current internet simplifies the allocation target of mobile social cloud resources in 5G/B5G, aiming at the interconnection environment of 5G/B5G everything, and supporting the application requirements of a 5G/B5G intelligent scene on huge connection, large flow, low time delay, high reliability and rapid evolution are higher and higher. The mobile social cloud is a resource sharing infrastructure, enables people with trust relationships to share computing or data services in a social circle (such as a community), is a means for enhancing multi-user collaboration, and can greatly promote resource exchange among participants.

Traditional social cloud computing and mobile cloud computing mostly adopt strategies with price as a leading factor, such as game theory, bidding auction and the like, during resource allocation so as to provide uniform and abstract infrastructure cloud service, and mobile social cloud computing needs to consider factors such as task computing amount, time delay, energy consumption and the like in addition to resource transaction price when facing computing and migrating computing service under an application scene. The invention meets the requirement of mobile social cloud resource allocation in a computing migration scene by a new reasonable and effective resource allocation mechanism, and is a more comprehensive and reasonable community-oriented mobile social cloud resource allocation mechanism compared with the existing method.

Disclosure of Invention

The invention aims to provide a mobile social cloud resource allocation technology in a 5G/B5G environment, which can meet the resource allocation requirement of a mobile social cloud under a computing migration scene according to a reasonable and effective resource allocation mechanism of factors such as time delay, energy consumption and social relations besides a resource transaction price.

The technical scheme of the invention is as follows: a multi-objective optimization resource allocation method based on decomposition and domination is characterized in that the goal of resource allocation is modeled from four dimensions respectively, the average time delay, the average energy consumption and the average user expense of a system are minimized and the average social benefit of a user is maximized by finding an optimal resource matching scheme, and the method mainly comprises the following steps:

step 1, determining a research scene of the method, and carrying out abstract modeling on a mobile social cloud system facing a community;

step 2, determining an optimization target for optimizing the resource allocation method;

step 3, modeling the resource allocation problem in the mobile social cloud environment facing the community from four dimensions;

step 4, constructing sub-objective functions through four dimensions, and establishing a problem model;

and 5, constructing a cloud resource allocation method for optimizing multiple targets based on the decomposition and domination relations.

The problem research scenario in the step 1 is that under the 5G/B5G environment, the calculation migration in the community is faced, and after the mobile social cloud system facing the community is abstracted and simplified, the problem research scenario can be roughly divided into three layers: resource domain, physical domain, and social domain.

The optimization goal in the step 2 is mainly considered from four dimensions, and the average time delay of the system, the average energy consumption and the average user expense are minimized and the average social benefit of the user is maximized by finding an optimal resource matching scheme.

The specific steps of the step 3 comprise:

step 3.1, when the demand side performs calculation migration, the time loss of the intermediate link is inevitably faced, and the considered range of the system time delay includes transmission (sending/receiving) time delay, queue waiting time delay, agent execution time delay, intermediate communication, result return time delay and the like. Because the calculation needing migration is usually that the self calculation amount is relatively large, and the result data amount is relatively small, the result return delay can be ignored, and the total delay characteristic vector is constructed as shown in the following formula (1):

wherein the content of the first and second substances,for the transmission delay of user i at time t,waiting and executing time delay for the user i at the time t;

step 3.2 the energy consumption of the computational migration is concentrated in the computational transfer process and the agent execution process. Because the social cloud uses an asymmetric transmission mode, and in the asymmetric transmission mode, the uplink transmission rate is much lower than the downlink transmission rate, the energy consumption evaluation is divided into two blocks, and the constructed energy consumption vector is as shown in formula (2):

wherein the content of the first and second substances,represents the transmission energy consumption, and is defined as shown in formula (3):

P_iis the transmission power, furthermoreRepresenting the agent execution energy consumption, considering two user-level modes of migration, i.e. single agent migration and multi-agent migration, where the energy consumption is divided into two cases, single agent mode such as formula (4) and multi-agent mode such as formula (5), accordingly.

Wherein the content of the first and second substances,andrespectively representing the CPUcycles consumed by the agent j when executing 1bit code, the code scale of the idx calculation migrated to the agent j by the demander i and the energy value consumed by the agent j when running 1 CPUcycle;

step 3.3 virtual social Currency Cost model is intended to evaluate social virtual Currency costs according to user transaction amounts, and construct a virtual currency Cost spent by a requester i in a transaction at time t_i,tAs shown in formula (6):

wherein the content of the first and second substances,is cost price, Com_i,idx,j,tAnd the unit of the calculation amount of the idx calculation in the self to-be-migrated calculation that the demand side i migrates to the agent j at the time t is CPUcycle.

Step 3.4, social benefit is social harvest brought to a demand party by a certain agent selection scheme, and is a measure for fully exerting the social relationship in the MSN, and a social benefit vector is constructed as shown in a formula (7):

wherein, b_iIs a weight coefficient, SR is social reputation, IntimD is intimacy, representing the strength of the social relationship between the social user and the agent, with a value range of [0,1]。

The specific steps of the step 4 comprise:

step 4.1, according to the task priority queue shown in the formula (8), determining that the system sub-target is average time delay minimization, and the objective function is shown in the formula (9):

step 4.2, determining the average energy consumption minimization objective function of the system sub-target users as formula (10):

step 4.3, determining the system sub-target user average virtual currency cost minimization objective function as (11):

step 4.4, determining the social average benefit maximization objective function of all users of the system sub-targets as formula (12):

in order to maintain consistency in the target optimization direction, the reciprocal of the above formula is converted into the formula (13):

step 4.5, by searching an optimal resource matching scheme, minimizing the average time delay of the system, the average energy consumption and the average user expense and maximizing the average social benefit of the user, the problem model target f (x) can be expressed as formula (14), formula (15) and formula (16):

minF(X)＝(f₁(X),...,f_m(X))^T (14)

X＝(x₁,...x_N)^T (15)

m＝4 (16)

wherein f is₁Representing the mean value of energy consumption of the user, f₂Representing the user's virtual social currency payout mean, f₃Representing the mean value of the user delay, f₄And the reciprocal of the social benefit mean value of the user is represented, X is a scheme set for resource allocation in a calculation migration scene, and m is the target number.

The specific steps of the step 5 are as follows:

step 5.1, converting multi-target problems such as time delay, energy consumption, social benefits, virtual social expenditure and the like in each resource allocation scheme into single-target problems which are easier to solve and understand through decomposition;

step 5.2, scheme sets are sorted quickly and non-dominantly, and Pareto domination hierarchical relations of the whole feasible scheme system are determined according to multiple targets such as time delay, energy consumption, social benefits and virtual social expenditure;

step 5.3, the non-dominated sorting aims at the feasible solutions in the solution set, namely feasible solutions, and introduces constraint violation values for the infeasible solutions to process the problem of the infeasible solutions.

The invention has the main beneficial effects that: the mobile social network which can be evolved by utilizing the real social contact and online transaction of the community has the natural advantages of solving trust and accountability, so that the mobile social cloud facing the community provides a resource sharing and agent computing service platform based on relationship and trust for community users under the environment of 5G/B5G by means of the social network, the computing capability of equipment with limited performance is expanded, and the utilization rate of equipment with surplus resources can be maintained at a relatively high level. Compared with the traditional calculation migration and resource allocation, the method introduces factors such as time delay and energy consumption of the calculation migration and social factors in the mobile social network, considers the problems of overhead and energy consumption and time delay caused by migration after resource allocation, and better exerts social interaction to form a scientific, reasonable and comprehensive mobile social cloud resource allocation mechanism, develops a prospect for the mobile terminal to run complex application under the conditions of insufficient computing power, electric quantity and the like, meanwhile, the approach of the devices can avoid the problems of network delay and jitter caused by transferring the calculation to a public cloud or a remote cloud, is a powerful attempt for distributing the calculation force to the edge, particularly the problem of high concentration of mobile devices and good utilization effect of the devices like a community with dense population, the method also has important value for constructing a novel 21 st century informatization community with resource intercommunication and computing power intercommunication. It is worth pointing out that the method has strong potential of scene transplantation and expansion, and can be popularized to personnel and equipment intensive areas such as schools, industrial parks and even mobile communities in the future.

Drawings

FIG. 1 is a flow chart of multi-objective optimized resource allocation according to the present invention.

FIG. 2 is a system abstraction model of the present invention.

FIG. 3 is a system framework diagram of the present invention.

FIG. 4 is a community-oriented mobile social network of the present invention.

FIG. 5 is a transaction module of the present invention.

FIG. 6 is a system monitoring module of the present invention.

FIG. 7 is a schematic representation of the PBI method of the present invention.

FIG. 8 impact of different resource allocation policies on the system and transactions.

Detailed Description

The invention provides a mobile social cloud resource allocation technology in a 5G/B5G environment, and key steps related to the method are described in detail below with reference to the accompanying drawings 1-8.

A mobile social cloud resource allocation technique in a 5G/B5G environment, comprising the following aspects: the method comprises the steps of abstracting a system model, constructing a system framework, decomposing and dominating based multi-objective optimization resource allocation algorithm and case evaluation.

The system abstraction model mainly comprises the following parts:

1.1 resource Domain

The resource domain characterizes the resource usage of the user equipment, and here, only the CPU and the stored resource information are listed, and in addition, the resource domain may also include network bandwidth, battery remaining capacity, and the like. Generally, the resource situation of the party sending the migration request is not very optimistic, such as slow calculation rate and insufficient battery remaining power; in contrast, the party that publishes the resource information is typically relatively resource-rich in its own right.

1.2 physical Domain

The physical domain characterizes the sharing of resources and migration of computing services among devices within the community.

1.3 social Domain

The social domain reflects social topology among users in the community, and the strength of the social relationship of each pair of users in the social topology has certain difference, namely high or low. The computing migration of the physical domain is performed based on a social relationship graph in the social domain, the migration generally occurs between two parties with social relationships, resource allocation behaviors based on migration computing in the resource domain are triggered, and resources such as a CPU (central processing unit), a storage and the like are allocated in advance by adopting an appropriate strategy, so that resource requirements and constraint conditions (such as time delay constraint and the like which need to be considered during computing migration) required by remotely executing computing tasks after migration are met.

2. Construction system frame

From the functional module, the system framework can be roughly divided into the following parts: the system comprises a social network management module, a transaction module, a monitoring module, an incentive module and the like, wherein user preferences (such as a migration mode, a transaction object range and the like) and a request queue, a resource pool and the like are involved in the middle. The social network management module is responsible for maintaining a social relationship graph, such as obtaining social topology information from a system, updating the topology graph timely and the like; the transaction module is responsible for resource matching and transaction settlement; the incentive module is responsible for maintaining the virtuous and sustainable quality of the overall transaction quality of the social cloud market by operating an efficient incentive and punishment strategy, such as a user who pertinently and severely attacks the malicious behavior of the user and actively contributes to the appreciation; the monitoring module is responsible for perceiving user behaviors, tasks, resource conditions and the like in the mobile social cloud market; the service evaluation module is responsible for evaluating the service providing level of the agent party;

the system framework mainly comprises the following parts:

2.1 social network management Module

The mobile social network facing the community is a network organization form generated by community users with certain social relations, and may be developed based on real social relations or generated through transaction behaviors. The main work of the social network management module is to maintain a logical graph of social relationships, including adding and removing edges in a social topological relationship network, wherein adding is mainly through adding friends and other ways in the real social relationships in the community, and removing is an action automatically taken by a system in which the social relationship strength of two parties is reduced to zero due to attenuation or punishment. In addition, considering the persistent form (i.e. text or database storage form) of the social topology of the system, the formatted data needs to be extracted and parsed by means of the social adapter.

2.2 transaction Module

The transaction matching process processed by the transaction module and some operations directly related after completion are mainly divided into two parts: a resource allocation submodule and a settlement submodule.

The user broadcasts own calculation request, the calculation request is selectively and uniformly placed in a request queue after being monitored and filtered, the agent publishes own available resource information in a broadcasting mode, the resource information is selectively added into a resource pool by a system after being monitored and filtered, an optimal allocation scheme is decided by operating a resource allocation algorithm, and both trading parties based on the scheme complete trading behaviors by settling social currency and providing agent calculation service.

2.3 monitoring Module

In the mobile social cloud market, various behaviors exist at any moment, including the fact that a user or an agent issues a request or resource information, the fact that the user provides service for the agent is evaluated and the like, and in order to ensure that the cloud market operates normally and immediately, various sensors (sensors) are arranged, wherein the sensors comprise a user Sensor, a calculation Sensor, a resource Sensor and the like, are triggered by transactions, such as issuing, settlement, evaluation and other transactions, and are responsible for updating the state or attribute value of a related object, such as adding the user request to a request queue, filtering illegal requests, recording the issuing and changing information of tasks or hardware resources and the like.

In addition, in consideration of the problem of cost increase of communication with other users in the community caused by base station switching when the base stations are crossed, the monitoring module is also responsible for judging the geographic positions of the users so as to judge whether the cloud users of the community initiate a computing migration request or agent computing service messages in the legal range around the community.

2.4 service evaluation Module

In a community-oriented mobile social cloud, the end of each round of transaction eventually occurs with the act of service scoring of agents providing agent computing by users as migration computing. By this scoring of the agents by the user, the quality of service offerings of the agents, i.e., the degree of fulfillment of the SLA agreements, can be well assessed. The service evaluation relates to the hierarchical division of the evaluation and the discrimination of the system on the true evaluation and the malicious evaluation.

2.5 excitation Module

The performance and reliability of a social cloud trading system relies on users 'cooperation in sharing computing power, and since each user is rational and selfish, in order to change the user's selfish psychology, strengthen true cooperation with each other and encourage trading, an appropriate marketing mechanism is needed to encourage user participation and allocate resources among multiple users, and thus an incentive mechanism is introduced.

The incentive mechanism is mainly embodied in the moderate increase of the social reputation of the agent and the intimacy of the two parties on the front. The higher the reputation of the user, the greater the probability that the next round will be selected as an agent or request to compute the migration success. The mechanism fully considers the performance of the agent in the current round of transaction, and specifically comprises contribution degree, popularity degree, user evaluation feedback and the like. On the other hand, the incentive mechanism also increases the positivity of the user to participate in the transaction and comply with the SLA agreement from the negative side through penalty measures. The method realizes the constraint on the standardization of transaction behaviors and the incentive for providing better service level for the agent and improves the enthusiasm of more community users for contributing own equipment resources to the agent through some incentive variables, such as the contribution or the selected proportion of the agent, the self credit of the agent, the evaluation of the user on the agent service level and the like. This also covers the system's graded penalty measures for unintentional or malicious users. The punishment modes in the current incentive mechanism are also various, the punishment frightening effect is fully played through research, the transaction peak period of the agent on the day is predicted by capturing the transaction amount time sequence of the agent on the previous day, a punishment time window (transaction forbidding) at the peak moment is set, and the data used for prediction can be updated in time, so that malicious behaviors are fully attacked. In addition, an instant penalty window is additionally set, so that a two-stage penalty system is formed.

3. Multi-objective optimization resource allocation algorithm based on decomposition and domination

In order to convert multi-objective problems such as time delay, energy consumption, social benefits, virtual social expenditure and the like in each resource allocation scheme into a single-objective problem which is easier to solve and understand, the resource matching in the migration calculation scene adopts a multi-objective optimization resource allocation algorithm CMOEA/DD algorithm based on decomposition and domination, and the algorithm is established based on the decomposition and domination relationship. Meanwhile, the grade hierarchical relationship of each scheme in the scheme set can be established through the non-dominated relationship, and the optimization direction of the next scheme can be better grasped by means of the grade hierarchy of the current scheme set in the process of solving the optimal scheme.

3.1 decomposition

The decomposition method adopted by the CMOEA/DD algorithm is a penalty-based boundary intersection (PBI), as shown in formulas (17) - (22), see fig. 7. :

minimize g^pbi(x|w,z^*)＝d₁+θd₂ (17)

s.t.x∈Ω (18)

wherein z is^*Is the target vector of the ideal solution, d₁Is the current solution scheme and z^*Distance of d₂Is the current solution scheme and z^*Is a user-defined non-negative real number, representing a penalty factor. w is the weight vector and m is the number of targets contained in the scheme. g^pbiThe function solves the penalty value brought by the deviation of a certain scheme and an ideal scheme solution in a mode of converting multiple targets such as time delay, energy consumption, social benefits and virtual social expenditure into a single target. The punishment value reflects the deviation degree of the current solution and the ideal solution in the solving process, namely the deviation degree of the ideal optimal evolution direction and the actual evolution direction, g^pbiThe larger the function value is, the larger the deviation from the optimal solution is, and the larger the punishment degree is.

3.2 fast non-dominated ordering of solution sets

In Pareto dominance relationships, for minimizing multiple objectives: f. of₁(x),...,f_m(x)，x_u,x_vE.g. omega. If it isAll have f_i(x_u)≤f_i(x_v) And at least one j, j ═ 1,. multidot.m } is present, satisfying f_i(x_u)≤f_i(x_v) Then x_uDominating x_vIs marked as x_u＜x_v. The fast non-dominated sorting of the scheme set is to determine Pareto dominated hierarchical relationship of the whole feasible scheme system according to multiple targets such as time delay, energy consumption, social benefits and virtual social expenditure, and algorithm pseudo codes of the Pareto dominated hierarchical relationship are as shown in a table 1:

table 1 scheme set fast non-dominated sorting algorithm

3.3 constraint handling

It should be noted that the non-dominated sorting is directed to feasible solutions in the solution set, namely feasible solutions (feasibilities), but the CMOEA/DD evolution algorithm allows the feasible solutions and the infeasisibilities to coexist. For this problem, CMOEA/DD introduces a constraint violation value (CV) to handle the problem of non-feasible solutions, as in equation (23), where the parenthetical operator < α > returns the absolute value of α if α <0, and 0 otherwise. It is clear that the smaller CV (x), the better the quality of x, and the CV for a feasible solution is always 0.

The diversity and openness of the knowledge is retained to some extent by the CV method. Since the equality constraint is constantly equal to 0, only the inequality constraint needs to be considered in the CV value solving process.

4. Evaluation of example

This patent has carried out test experiment on mobile social network blogCatalog's data set.

4.1 evaluation index

Four performance evaluation indexes are adopted, which are respectively as follows: average energy consumption index, average time delay index, average expenditure index and average social profit index.

By adopting a traditional Reverse Auction Algorithm (RAA) of the social cloud, when a user puts forward a calculation migration request, a plurality of agents bid in sequence, and finally bid by a person with a lower price to check the RAA and the method (MSCRAM) of the invention.

4.2 evaluation results

On the data set BlogCatalog, a method test is performed from the average energy consumption index, the average time delay index, the average expenditure index and the average social profit index, and the test result please refer to the test result of fig. 8.

Compared with the traditional experiment results of the RAA resource allocation strategy and the MSCRAM allocation strategy which take price as the guide from the aspect of the allocation strategy, the method of summing the RAA resource allocation strategy and the MSCRAM resource allocation strategy and calculating the occupied proportion of each index value of the result obtained by the RAA resource allocation strategy and the MSCRAM resource allocation strategy respectively is adopted.

Besides price factor reverse auction strategy wins, the mechanism provided by the invention is more advantageous in other aspects such as system average time delay, system average energy consumption and average social benefit, and the like, thereby illustrating the beneficial effect of the invention on the comprehensive performance of cloud resource allocation.