CN105208076B - A kind of multiple target service combining method perceived based on correlation - Google Patents

本发明涉及一种基于相关性的多目标服务组合方法，包括：将候选服务集合C_i中没有服务质量相关性的候选服务存入相匹配的第一候选服务集合对中的候选服务两两比较以获取优胜候选服务并存入优胜候选服务集合从C_i中删除相对应的非优胜候选服务以获取相匹配的子服务集合C_i′；将所有C_i′组合以形成新的服务组合解空间S′；从S′中随机选取多个服务组合解形成代表解集合计算中每个服务组合解S_p的粗略服务质量值并进行分层；穷举所选择的前s层中全部服务组合解的相关性信息，以获取全部该服务组合解的实际服务质量值；根据实际服务质量值对所对应的服务组合解进行排序，选择前K个服务组合解以获取次优服务组合解集合。本发明能够快速得到次优解，提高求解效率。

The present invention relates to a correlation-based multi-objective service composition method, comprising: storing candidate services without service quality correlation in a candidate service set C _i into a matching first candidate service set right Compare the candidate services in pairwise to obtain the winning candidate service and store it in the winning candidate service set Delete the corresponding non-winning candidate services from C _i to obtain the matching sub-service set C _i ′; combine all C _i ′ to form a new service combination solution space S′; randomly select multiple services from S′ Combining solutions to form representative solution sets calculate The rough service quality value of each service combination solution S _p in each service combination solution is stratified; exhaustively enumerates the correlation information of all service combination solutions in the selected first s layer to obtain the actual service quality value of all the service combination solutions; according to The actual service quality values sort the corresponding service combination solutions, and select the top K service combination solutions to obtain the suboptimal service combination solution set. The invention can quickly obtain the suboptimal solution and improve the solution efficiency.

一种基于相关性感知的多目标服务组合方法A multi-objective service composition method based on correlation awareness

技术领域technical field

本发明涉及云计算管理与控制技术领域，尤其涉及一种基于相关性感知的多目标服务组合方法。The invention relates to the technical field of cloud computing management and control, in particular to a method for combining multi-target services based on correlation perception.

背景技术Background technique

随着SOA(Service-Oriented Architecture，面向服务的体系结构)技术和web服务技术的出现，互联网上出现了越来越多的具有封装接口的服务应用。随着互联网上应用个数的急剧增多，出现了很多的服务工业标准，例如：WSDL(Web Services DescriptionLanguage，网络服务描述语言)和UDDI(Universal Description Discovery andIntegration，通用描述、发现与集成服务)等。商业过程通常需要一系列的子服务进行协作，以完成需要的功能，这个过程叫做服务组合。服务组合使得开发者可以根据预先定义好的需求将子服务组合成为一个工作流。With the emergence of SOA (Service-Oriented Architecture, service-oriented architecture) technology and web service technology, more and more service applications with encapsulated interfaces appear on the Internet. With the rapid increase in the number of applications on the Internet, many service industry standards have emerged, such as: WSDL (Web Services Description Language, Network Service Description Language) and UDDI (Universal Description Discovery and Integration, general description, discovery and integration services) and so on. Business processes usually require a series of sub-services to cooperate to complete the required functions. This process is called service composition. Service composition enables developers to compose sub-services into a workflow based on predefined requirements.

不同供应商提供功能相同或者相近悬念具有不同非功能特征的候选服务。非功能特征主要以服务质量来体现，当一个服务请求到达时，由于服务质量具有多维的属性，例如包括有响应时间、吞吐率和可用性等，是一个多目标优化的问题，因此如何选择合适的子服务以实现最优的端到端服务质量成为研究热点。Different suppliers provide candidate services with the same or similar functions but with different non-functional characteristics. The non-functional characteristics are mainly reflected by the quality of service. When a service request arrives, since the quality of service has multi-dimensional attributes, such as response time, throughput and availability, etc., it is a multi-objective optimization problem, so how to choose the appropriate Sub-services to achieve optimal end-to-end service quality has become a research hotspot.

为了处理不同服务质量属性之间的权衡和折中，现有技术中将多目标优化的问题转化为单目标优化的问题，主要分为线性加权和将目标转换为约束条件两类。其中，线性加权将不同目标归一化，然后设置相应的权重再相加。一方面，在归一化过程中需要知道目标的最大值、最小值或者平均值，然而实际应用中不容易获取这些值；另一方面，设置权重还需要知道不同目标的优先级，实际应用中优先级数据也不容易知道，且如何设置约束条件也未解决。In order to deal with the trade-offs and compromises between different service quality attributes, the prior art converts the problem of multi-objective optimization into a single-objective optimization problem, which is mainly divided into two categories: linear weighting and converting objectives into constraints. Among them, the linear weighting normalizes different targets, and then sets the corresponding weights and adds them together. On the one hand, it is necessary to know the maximum value, minimum value or average value of the target in the normalization process, but it is not easy to obtain these values in practical applications; on the other hand, setting the weight also needs to know the priority of different targets, in practical applications Priority data is also not easily known, and how to set constraints is also unresolved.

并且，现有技术中很多已有的服务组合解决方案没有考虑服务之间的服务质量相关性，实际应用时一个子服务的服务质量值可能会依赖其他的子服务。例如：选择将两个或者多个某公司的子服务放置在同一个工作流中，该两个或者多个某公司的子服务的服务质量可以打折。又如，航空预定公司在收费过程中，若用户使用信用卡支付，可能会收多余的费用；若用户使用借记卡支付，则无需多余费用。再如，选择将两个子服务放置在同一台服务器上，两者之间传输时间将极大减小，组合服务的响应时间也相应减小。若不考虑服务质量的相关性则会影响得到服务组合的服务解；但将服务质量相关性考虑进服务组合，还会使得服务组合问题变得非常复杂。因此，如何提高求解效率，是亟需解决的技术问题。Moreover, many existing service composition solutions in the prior art do not consider the service quality correlation between services, and the service quality value of a sub-service may depend on other sub-services in actual application. For example, if you choose to place two or more sub-services of a certain company in the same workflow, the service quality of the two or more sub-services of a certain company can be discounted. As another example, during the charging process of an airline reservation company, if the user pays with a credit card, the extra fee may be charged; if the user pays with a debit card, no extra fee is required. For another example, if you choose to place two sub-services on the same server, the transmission time between the two will be greatly reduced, and the response time of the composite service will also be reduced accordingly. If the correlation of service quality is not considered, the service solution of service composition will be affected; but if the correlation of service quality is taken into account in service composition, the problem of service composition will become very complicated. Therefore, how to improve the solution efficiency is an urgent technical problem to be solved.

发明内容Contents of the invention

本发明的目的之一在于提供一种基于相关性的多目标服务组合方法，以提供一种求解效率高、考虑相关性的多目标服务组合方法。One of the objectives of the present invention is to provide a multi-objective service composition method based on correlation, so as to provide a multi-objective service composition method with high solution efficiency and consideration of correlation.

为实现上述目的，本发明提出了一种基于相关性的多目标服务组合方法，包括：In order to achieve the above object, the present invention proposes a multi-objective service combination method based on correlation, including:

候选服务集合C_i中每个候选服务包含多个属性，若任选一个候选服务与剩余候选服务中至少一个候选服务相对应的属性具有相关性，则所对应的候选服务具有服务质量相关性；若一个候选服务与所有其他候选服务的相对应的属性都没有相关性，则该候选服务没有服务质量相关性，并将该没有服务质量相关性的候选服务存入相匹配的第一候选服务集合 Each candidate service in the candidate service set C _i includes a plurality of attributes, if one of the candidate services is selected to have a correlation with at least one of the remaining candidate services, the corresponding attribute has a correlation with the quality of service; If a candidate service has no correlation with the corresponding attributes of all other candidate services, the candidate service has no service quality correlation, and the candidate service without service quality correlation is stored in the matching first candidate service set

对所述第一候选服务集合中的候选服务进行两两比较以获取优胜候选服务与非优胜候选服务，将所获取的优胜候选服务存入优胜候选服务集合并从所述第一候选服务集合删除非优胜候选服务；For the first set of candidate services Compare the candidate services in pairwise to obtain the winning candidate service and the non-winning candidate service, and store the obtained winning candidate service in the winning candidate service set and from the first candidate service set Deletion of non-winning candidate services;

从所述候选服务集合C_i中删除相对应的非优胜候选服务以获取相匹配的子服务集合C′_i；Deleting the corresponding non-winning candidate service from the candidate service set C _i to obtain a matching sub-service set C'_i;

将所有子服务集合C′_i组合以形成新的服务组合解空间S′；从所述服务组合解空间S′中随机选取多个服务组合解形成代表解集合计算所述代表解集合中每个服务组合解S_p的粗略服务质量值并进行分层；Combine all sub-service sets C' _i to form a new service combination solution space S'; randomly select multiple service combination solutions from the service combination solution space S' to form a representative solution set Calculate the set of representative solutions Solve the rough service quality value of S _p for each service combination in and stratify;

穷举所选择的前s层中全部服务组合解的相关性信息，以获取全部该服务组合解的实际服务质量值；Exhaustively enumerate the correlation information of all service combination solutions in the selected first s layers to obtain the actual service quality values of all the service combination solutions;

根据实际服务质量值对所对应的服务组合解进行排序，选择前K个服务组合解以获取次优服务组合解集合。According to the actual service quality value, the corresponding service combination solutions are sorted, and the top K service combination solutions are selected to obtain the suboptimal service combination solution set.

可选地，所述候选服务集合C_i中每个候选服务包含多个属性，若任选一个候选服务与剩余候选服务中至少一个候选服务相对应的属性具有相关性，则所对应的候选服务具有服务质量相关性；若一个候选服务与所有其他候选服务的相对应的属性都没有相关性，则该候选服务没有服务质量相关性，并将该没有服务质量相关性的候选服务存入相匹配的第一候选服务集合的步骤中采用如下公式判断每个候选服务是否具有服务质量相关性：Optionally, each candidate service in the set of candidate services C _i includes a plurality of attributes, and if any one of the candidate services has a correlation with at least one attribute corresponding to the remaining candidate services, then the corresponding candidate service Has quality of service correlation; if a candidate service has no correlation with the corresponding attributes of all other candidate services, the candidate service has no quality of service correlation, and the candidate service without quality of service correlation is stored in the matching The set of first candidate services for In the step, the following formula is used to judge whether each candidate service has service quality correlation:

式中，表示用于完成子服务i的候选服务，代表候选服务的第r个属性是否具有相关性；代表候选服务是否具有服务质量相关性；“∨”代表求并运算。In the formula, denotes a candidate service for completing subservice i, representative candidate service Whether the rth attribute of is relevant; representative candidate service Whether it has quality of service correlation; "∨" represents a union operation.

可选地，所述候选服务集合C_i中每个候选服务包含多个属性，若任选一个候选服务与剩余候选服务中至少一个候选服务相对应的属性具有相关性，则所对应的候选服务具有服务质量相关性；若一个候选服务与所有其他候选服务的相对应的属性都没有相关性，则该候选服务没有服务质量相关性，并将该没有服务质量相关性的候选服务存入相匹配的第一候选服务集合的步骤之后，还包括：Optionally, each candidate service in the set of candidate services C _i includes a plurality of attributes, and if any one of the candidate services has a correlation with at least one attribute corresponding to the remaining candidate services, then the corresponding candidate service Has quality of service correlation; if a candidate service has no correlation with the corresponding attributes of all other candidate services, the candidate service has no quality of service correlation, and the candidate service without quality of service correlation is stored in the matching The set of first candidate services for After the steps, also include:

计算每个候选服务集合中各个候选服务的Grade值，根据Grade值对所述候选服务集合中全部候选服务进行升序排序。Compute each set of candidate services The Grade value of each candidate service in , and set the candidate services according to the Grade value All candidate services are sorted in ascending order.

可选地，所述从所述候选服务集合C_i中删除相对应的非优胜候选服务以获取相匹配的子服务集合C′_i的步骤中根据如下公式得到子服务集合C′_i：Optionally, in the step of deleting the corresponding non-winning candidate service from the candidate service set C _i to obtain the matching sub-service set C' _i , the sub-service set C' _i is obtained according to the following formula:

式中，C′_i为子服务集合，C_i为候选服务集合，为优胜候选服务集合，为候选服务集合。In the formula, C′ _i is the sub-service set, C _i is the candidate service set, Service collection for the winning candidate, A collection of candidate services.

可选地，所述将全部子服务集合C′_i组合以形成新的服务组合解空间S′的步骤中采用如下公式获取服务组合解空间S′：Optionally, in the step of combining all sub-service sets C' _i to form a new service combination solution space S', the following formula is used to obtain the service combination solution space S':

S′＝C′₁×C′₂×…×C′_m S′=C′ ₁ ×C′ ₂ ×…×C′ _m

式中，C′_i代表子服务集合，i＝1、2、3、……、m，符号“×”代表笛卡尔乘积。In the formula, C′ _i represents a set of sub-services, i=1, 2, 3, ..., m, and the symbol "×" represents a Cartesian product.

可选地，所述从所述服务组合解空间S′中随机选取多个服务组合解形成代表解集合计算所述代表解集合中每个服务组合解S_p的粗略服务质量值并进行分层的步骤包括：Optionally, randomly selecting a plurality of service combination solutions from the service combination solution space S′ to form a representative solution set Calculate the set of representative solutions The steps of solving the rough quality of service value of S _p for each service combination in and performing stratification include:

选取所述服务组合解空间S′中任意两个服务组合解S_pi和S_pj进行两两比较，若服务组合解S_pi优胜服务组合解S_pj，将服务组合解S_pj加入到所述服务组合解S_pi的队列中，且将服务组合解S_pi加入所述服务组合解S_pj的队列中；Select any two service combination solutions S _pi and S _pj in the service combination solution space S′ for pairwise comparison, if the service combination solution S _pi is superior to the service combination solution S _pj , add the service combination solution S _pj to the service combination solution Queue for combinatorial solution S _pi , and add the service combination solution S _pi to the queue of the service combination solution S _pj middle;

定义利用L_i代表第i层的服务组合解集合；从第一层i＝1开始，初始化当时，找到所有的服务组合解S_pj，并且将该S_pj加入L_i中，然后找到队列中所有的服务组合解从每一个队列中删除服务组合解S_pj，再将服务组合解从RemainingSet中删除；definition Utilize L _i to represent the solution collection of the i-th layer service combination; start from the first layer i=1, initialize when , find all The service combination solution S _pj , and add this S _pj to L _i , and then find the queue All service composition solutions in from each queue Delete the service composition solution S _pj in the service composition solution, and then Delete from RemainingSet;

重复上述过程，直至从而获取服务组合解空间S′共有q层服务组合解；Repeat the above process until In order to obtain the service composition solution space S′, there are q layers of service composition solutions;

其中，队列是所有可以优胜服务组合解S_p的服务组合解的集合，是所有S_p优胜的解的集合。Among them, the queue is the set of all service combination solutions that can win the service combination solution S _p , is the set of all S _p winning solutions.

本发明实施例通过计算服务组合解的粗略服务质量进行分层，并计算所选择对应层中候选服务的相关性信息，以获取实际服务质量值，从而能够处理实际系统带有相关性的服务组合情况，能够快速得到实际系统的次优解，从而提高求解效率。The embodiment of the present invention performs layering by calculating the rough service quality of the service composition solution, and calculates the correlation information of the candidate services in the selected corresponding layer to obtain the actual service quality value, so as to be able to process the service composition with correlation in the actual system In this case, the suboptimal solution of the actual system can be quickly obtained, thereby improving the solution efficiency.

附图说明Description of drawings

通过参考附图会更加清楚的理解本发明的特征和优点，附图是示意性的而不应理解为对本发明进行任何限制，在附图中：The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the invention in any way. In the accompanying drawings:

图1是本发明实施例提供的一种基于相关性的多目标服务组合方法框图；FIG. 1 is a block diagram of a correlation-based multi-target service combination method provided by an embodiment of the present invention;

图2是本发明一实施例中对候选服务优胜比较流程示意图；FIG. 2 is a schematic diagram of a comparison flow for candidate services in an embodiment of the present invention;

图3是本发明一实施例中求解服务组合解流程示意图。Fig. 3 is a schematic diagram of a process for solving service combination solutions in an embodiment of the present invention.

具体实施方式Detailed ways

下面结合附图和实施例，对本发明的具体实施方式作进一步详细描述。以下实施例用于说明本发明，但不用来限制本发明的范围。The specific implementation manners of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

本发明实施例提供了基于相关性的多目标服务组合方法，如图1所示，包括：The embodiment of the present invention provides a multi-target service combination method based on correlation, as shown in Figure 1, including:

S100、候选服务集合C_i中每个候选服务包含多个属性，若任选一个候选服务与剩余候选服务中至少一个候选服务相对应的属性具有相关性，则所对应的候选服务具有服务质量相关性；若一个候选服务与所有其他候选服务的相对应的属性都没有相关性，则该候选服务没有服务质量相关性，并将该没有服务质量相关性的候选服务存入相匹配的第一候选服务集合 S100. Each candidate service in the candidate service set C _i contains a plurality of attributes. If any candidate service has a correlation with at least one attribute corresponding to the candidate service in the remaining candidate services, the corresponding candidate service has a quality of service correlation. If a candidate service has no correlation with the corresponding attributes of all other candidate services, then the candidate service has no quality of service correlation, and the candidate service without quality of service correlation is stored in the matching first candidate collection of services

S200、对第一候选服务集合中的候选服务进行两两比较以获取优胜候选服务与非优胜候选服务，将所获取的优胜候选服务存入优胜候选服务集合并从第一候选服务集合删除非优胜候选服务；S200. Set the first candidate service Compare the candidate services in pairwise to obtain the winning candidate service and the non-winning candidate service, and store the obtained winning candidate service in the winning candidate service set and from the first candidate service set Deletion of non-winning candidate services;

S300、从候选服务集合C_i中删除相对应的非优胜候选服务以获取相匹配的子服务集合C′_i；S300. Delete the corresponding non-winning candidate service from the candidate service set C _i to obtain a matching sub-service set C'_i;

S400、将所有子服务集合C′_i组合以形成新的服务组合解空间S′；从服务组合解空间S′中随机选取多个服务组合解形成代表解集合计算代表解集合中每个服务组合解S_p的粗略服务质量值并进行分层；S400. Combine all sub-service sets _C'i to form a new service combination solution space S'; randomly select multiple service combination solutions from the service combination solution space S' to form a representative solution set Compute the set of representative solutions Solve the rough service quality value of S _p for each service combination in and stratify;

S500、穷举所选择的前s层中全部服务组合解的相关性信息，以获取全部该服务组合解的实际服务质量值；S500. Exhaustively enumerate the correlation information of all the service combination solutions in the selected first s layers, so as to obtain the actual service quality values of all the service combination solutions;

S600、根据实际服务质量值对所对应的服务组合解进行排序，选择前K个服务组合解以获取次优服务组合解集合。S600. Sort the corresponding service combination solutions according to the actual service quality values, and select the top K service combination solutions to obtain a set of suboptimal service combination solutions.

下面结合附图及其实施例对本发明实施例所提供方法的步骤展开说明。The steps of the method provided by the embodiment of the present invention will be described below in conjunction with the accompanying drawings and the embodiments thereof.

本发明实施例中，用集合I＝{1,2,…,m}表示m个子服务，对于每个一个子服务i有n_i个候选服务可以完成。该n_i个候选服务用候选服务集合来表示；服务组合解表示某一个具体的服务组合解，其中表示用于完成子服务i的候选服务；S＝C₁×C₂×…×C_m表示解空间。a_r表示第r个服务质量的属性，所有的服务质量属性用向量A＝(a₁,a₂,…,a_l)来表示。从子服务的服务质量值到组合服务的服务质量值的聚集函数用表示，聚集函数有相加、相乘、最大化、最小化、并和交等。In the embodiment of the present invention, m sub-services are represented by a set I={1,2,...,m}, and there are n _i candidate services that can be completed for each sub-service i. The n _i candidate services use a set of candidate services to represent; service composition solution Represents a specific service combination solution, where Indicates candidate services for completing sub-service i; S=C ₁ ×C ₂ ×...×C _m represents the solution space. a _r represents the attribute of the rth service quality, and all service quality attributes are represented by vector A=(a ₁ ,a ₂ ,...,a _l ). Aggregate function from QoS values of subservices to QoS values of composite services using Indicates that aggregate functions include addition, multiplication, maximization, minimization, union, and intersection.

本发明实施例中用Dep_r＝{<S_sp,v_sp>}表示第r个服务质量相关性集合，其中S_sp代表具有相关性的子服务组合，v_sp代表该子组合的服务质量值。二元变量代表候选服务的第r个属性是否具有相关性；代表候选服务是否具有服务质量相关性；“∨”代表并运算；代表候选服务第r个属性的默认值。In the embodiment of the present invention, Dep _r ={<S _sp ,v _sp >} is used to represent the rth service quality correlation set, wherein S _sp represents a sub-service combination with correlation, and v _sp represents the service quality value of the sub-combination . binary variable representative candidate service Whether the rth attribute of is relevant; representative candidate service Whether it has service quality correlation; "∨" stands for operation; representative candidate service The default value for the rth attribute.

V(S_p)＝(v₁(S_p),…,v_r(S_p),…,v_l(S_p))用于表示组合服务S_p的服务质量值。对于任意两个组合服务，S_p和S′_p，若下式成立，则称为S_p＞S′_p，即S_p优胜S′_p。V(S _p )=(v ₁ (S _p ),...,v _r (S _p ),...,v _l (S _p )) is used to represent the service quality value of the composite service S _p . For any two combined services, S _p and S′ _p , if the following formula holds, it is called S _p > S′ _p , that is, S _p is better than S′ _p .

其中，≥表示优于或相等，＞表示严格优于。最优的服务组合解是解空间S中所有不被其他解优胜的解集，即最优解集合对于目标值为最大化的服务质量属性，将其服务质量值取反。Among them, ≥ means superior or equal, and > means strictly superior. The optimal service combination solution is all the solution sets in the solution space S that are not superior to other solutions, that is, the optimal solution set For the quality of service attribute whose target value is maximized, its quality of service value is reversed.

首先，介绍步骤S100、候选服务集合C_i中每个候选服务包含多个属性，若任选一个候选服务与剩余候选服务中至少一个候选服务相对应的属性具有相关性，则所对应的候选服务具有服务质量相关性；若一个候选服务与所有其他候选服务的相对应的属性都没有相关性，则该候选服务没有服务质量相关性，并将该没有服务质量相关性的候选服务存入相匹配的第一候选服务集合的步骤。First, step S100 is introduced. Each candidate service in the candidate service set C _i contains multiple attributes. If any candidate service is correlated with the attribute corresponding to at least one candidate service among the remaining candidate services, the corresponding candidate service Has quality of service correlation; if a candidate service has no correlation with the corresponding attributes of all other candidate services, the candidate service has no quality of service correlation, and the candidate service without quality of service correlation is stored in the matching The set of first candidate services for A step of.

候选服务集合C_i中包含多个候选服务，每个候选服务包含多个属性。若对于其中一个候选服务，如果该候选服务的其中一个属性与该候选服务集合C_i中其他至少一个候选服务的相对应的属性有相关性，则判断该候选服务具有服务质量相关性。若对于其中一个候选服务，如果该候选服务与剩余候选服务中的所有候选服务都没有相关性，则判断该候选服务没有相关性。The candidate service set C _i contains multiple candidate services, and each candidate service contains multiple attributes. If for one of the candidate services, if one of the attributes of the candidate service is correlated with the corresponding attribute of at least one other candidate service in the candidate service set C _i , then it is determined that the candidate service has a quality of service correlation. If for one of the candidate services, if the candidate service has no correlation with all the candidate services in the remaining candidate services, it is determined that the candidate service has no correlation.

本发明实施例中，利用公式(1)来判断一个候选服务是否具有服务质量相关性 In the embodiment of the present invention, formula (1) is used to judge whether a candidate service has quality of service correlation

其中，二元变量表示候选服务是否具有服务质量相关性。where the binary variable represent candidate services Whether there is a quality of service correlation.

针对每个候选服务集合C_i，找出其中所有的候选服务，并将满足条件的候选服务放入候选服务集合该第一候选服务集合采用下式表示：For each candidate service set C _i , find out all candidate services, and put the candidate services that meet the conditions into the candidate service set The first set of candidate services It is represented by the following formula:

其次，介绍S200、对第一候选服务集合中的候选服务进行两两比较以获取优胜候选服务与非优胜候选服务，将所获取的优胜候选服务存入优胜候选服务集合并从第一候选服务集合删除非优胜候选服务的步骤。Secondly, introduce S200, the first candidate service set Compare the candidate services in pairwise to obtain the winning candidate service and the non-winning candidate service, and store the obtained winning candidate service in the winning candidate service set and from the first candidate service set Steps to delete services that are not winning candidates.

根据每个候选服务的相对应的属性的默认值，利用下式计算第一候选服务集合中每个候选服务的Grade值：According to the default value of the corresponding attribute of each candidate service, use the following formula to calculate the first set of candidate services The Grade value of each candidate service in:

并根据Grade值对第一候选服务集合中的候选服务进行排序，本发明实施例中采用升序排序。选择Grade值最小的候选服务置于第一候选服务集合的第一个，其他候选服务依次放入至相应的位置。And set the first candidate service according to the Grade value The candidate services in are sorted in ascending order in the embodiment of the present invention. Select the candidate service with the smallest Grade value placed in the first set of candidate services first, other candidate services Put them in the corresponding position one by one.

然后从第一候选服务集合中获取优胜候选服务与非优胜候选服务。Then from the first candidate service set Obtain the winning candidate service and the non-winning candidate service in .

本发明实施例中，定义优胜候选服务集合表示第一候选服务集合中最优的候选服务集合。变量c_k代表第一候选服务集合中第k个候选服务，二元变量IsDominated(k)＝1表示候选服务c_k已经被优胜了。In the embodiment of the present invention, the winning candidate service set is defined Indicates the first set of candidate services The best set of candidate services in . The variable c _k represents the first set of candidate services In the kth candidate service, the binary variable IsDominated(k)=1 indicates that the candidate service c _k has been won.

如图2所示，初始化优胜候选服务集合针对所有的候选服务c_k，初始化二元变量IsDominated(k)＝0。从候选服务c₁开始比较，直至最后一个候选服务c_k。针对候选服务c_k的比较称为第k轮比较，过程如下：如果候选服务c_k优胜候选服务c_j，那么标记IsDominated(j)＝1，再继续将候选服务c_k和后面的候选服务比较；不然，依次将候选服务c_k和之后的候选服务c_j(j>k,IsDominated(j)＝0)比较。如果候选服务c_j优胜候选服务c_k，标记IsDominated(k)＝1，并且直接跳到第k+1轮比较；否则，如果候选服务c_k和候选服务c_j互相不优胜，则将候选服务c_k与和后面的候选服务比较。该比较过程一直继续，直到进行到第一候选服务集合中最后一轮候选服务的比较。As shown in Figure 2, initialize the set of winning candidate services For all candidate services c _k , initialize the binary variable IsDominated(k)=0. The comparison starts from the candidate service c ₁ to the last candidate service c _k . The comparison for the candidate service c _k is called the k-th round of comparison, and the process is as follows: if the candidate service c _k wins the candidate service c _j , then mark IsDominated(j)=1, and then continue to compare the candidate service c _k with the following candidate service Compare; otherwise, compare the candidate service c _k with the subsequent candidate service c _j (j>k, IsDominated(j)=0) in turn. If the candidate service c _j wins the candidate service c _k , mark IsDominated(k)=1, and skip directly to the k+1th round of comparison; otherwise, if the candidate service c _k and the candidate service c _j are not superior to each other, the candidate service _ck and subsequent candidate services Compare. This comparison process continues until the first set of candidate services is reached Comparison of the last round of candidate services in .

遍历所有候选服务的IsDominated(k)，如果IsDominated(k)＝0，则将候选服务c_k加入优胜候选服务集合即优胜候选服务集合中保存了第一候选服务集合中最优的候选服务。Traverse the IsDominated(k) of all candidate services, if IsDominated(k)=0, add the candidate service c _k to the winning candidate service set That is, the collection of winning candidate services The set of first candidate services is saved in The best candidate service in .

再次，介绍S300、从候选服务集合C_i中删除相对应的非优胜候选服务以获取相匹配的子服务集合C′_i的步骤。Again, the step of S300, deleting the corresponding non-winning candidate service from the candidate service set C _i to obtain the matching sub-service set C' _i is introduced.

如果IsDominated(k)＝1，则候选服务c_k为非优选候选服务。根据如下公式求解子服务集合C′_i：If IsDominated(k)=1, the candidate service c _k is a non-preferred candidate service. Solve the sub-service set C′ _i according to the following formula:

第四，介绍将多个子服务集合C′_i组合以形成新的服务组合解空间S′；从所述服务组合解空间S′中随机选取多个服务组合解形成代表解集合计算所述代表解集合中每个服务组合解S_p的粗略服务质量值并进行分层的步骤。Fourth, introduce the combination of multiple sub-service sets C' _i to form a new service combination solution space S'; randomly select multiple service combination solutions from the service combination solution space S' to form a representative solution set Calculate the set of representative solutions The step of solving the rough quality of service value of S _p for each service combination in and carrying out stratification.

采用如下公式(4)将多个子服务集合C′_i组合以获取服务组合解空间S′。The following formula (4) is used to combine multiple sub-service sets C' _i to obtain a service combination solution space S'.

S′＝C′₁×C′₂×…×C′_m (4)S′=C′ ₁ ×C′ ₂ ×…×C′ _m (4)

从获取服务组合解空间S′中随机选取多个服务组合解形成代表解集合本发明一实施例中随机选择多个代表解，例如一实施例中选择10000个代表解，针对这10000个代表解，先不考虑相关性信息，直接通过聚集函数计算每个服务组合解S_p的粗略服务质量值V_c(S_p)。Randomly select multiple service combination solutions from the obtained service combination solution space S′ to form a representative solution set In one embodiment of the present invention, a plurality of representative solutions are randomly selected. For example, in one embodiment, 10,000 representative solutions are selected. For these 10,000 representative solutions, the correlation information is not considered first, and each service combination solution S _p is directly calculated through an aggregation function. A rough quality of service value V _c (S _p ) for .

聚集函数下式(5)所示：The aggregate function is shown in the following formula (5):

粗略服务质量值V_c(S_p)如式(6)所示：The rough service quality value V _c (S _p ) is shown in formula (6):

V_c(S_p)＝(v_1c(S_p),…,v_rc(S_p),…,v_lc(S_p)) (6)V _c (S _p )＝(v _1c (S _p ),…,v _rc (S _p ),…,v _lc (S _p )) (6)

例如，对于聚集函数为相加的属性，组合服务解的服务质量值等于每个子服务的服务质量值之和；对于聚集函数为相乘的属性，组合服务解的服务质量值等于每个子服务的服务质量的乘积。For example, for an attribute whose aggregation function is addition, the QoS value of the combined service solution is equal to the sum of the QoS values of each sub-service; for an attribute whose aggregation function is multiplication, the QoS value of the combined service solution is equal to the QoS value of each sub-service The product of quality of service.

根据粗略服务质量值V_c(S_p)对所选择的10000个代表解进行分层和排序，根据代表解的分层以获取该10000个代表解的好坏信息。The selected 10,000 representative solutions are stratified and sorted according to the rough service quality value V _c (S _p ), and the quality information of the 10,000 representative solutions is obtained according to the stratification of the representative solutions.

第五，介绍S500、穷举所选择的前s层中全部服务组合解的相关性信息，以获取全部该服务组合解的实际服务质量值的步骤。Fifth, introduce S500, the step of exhaustively enumerating the correlation information of all service combination solutions in the selected first s layers to obtain the actual service quality values of all the service combination solutions.

根据上述步骤中每个服务组合解S_p的粗略服务质量值和一共的层数，确定所需要选择的服务组合解的层数s，例如需要前三层的服务组合解，则穷举该前三层服务组合解的所有的相关性信息，从而可以得到这些服务组合解的实际服务质量值，从而获取服务组合解空间共有q层服务组合解。According to the rough service quality value and the total number of layers of each service combination solution S _p in the above steps, determine the layer number s of the service combination solution that needs to be selected. All the correlation information of the three-layer service composition solution, so that the actual service quality value of these service composition solutions can be obtained, and the service composition solution space can be obtained There are q-level service composition solutions.

对所选择的前s层服务组合解进行分层过程如下：如图3所示，针对每一个服务组合解S_p，设置两个队列和其中，队列代表所有可以优胜服务组合解S_p的服务组合解的集合，队列代表所有服务组合解S_p优胜的服务组合解的集合。The layering process for the selected service combination solutions of the first s layers is as follows: As shown in Figure 3, for each service combination solution S _p , set up two queues and Among them, the queue Represents the collection of all service composition solutions that can win the service composition solution S _p , the queue Represents the collection of service composition solutions that all service composition solutions S _p win.

对于代表解集合中所有的服务组合解S_pi，初始化 其中，服务组合解S_pi表示代表解集合中第i个服务组合解。For the set of representative solutions All service combinations in S _pi , initialize Among them, the service combination solution S _pi represents the set of solutions The i-th service combination solution in .

从代表解集合中所有的服务组合解中任选两个服务组合解进行两两比较，若服务组合解优胜服务组合解S_pj，则将服务组合解S_pi加入服务组合解S_pj的队列中，将服务组合解S_pj加入服务组合解的队列中；若服务组合解S_pj优胜服务组合解S_pi，则将服务组合解S_pi加入服务组合解S_pj队列中，服务组合解S_pj加入服务组合解S_pi的队列中。Solve collections from representatives Among all the service composition solutions in , choose two service composition solutions for pairwise comparison, if the service composition solution The winning service combination solution S _pj , then add the service combination solution S _pi to the queue of the service combination solution S _pj , add the service composition solution S _pj into the service composition solution queue middle; if the service combination solution S _pj wins the service combination solution S _pi , then add the service combination solution S _pi to the service combination solution S _pj queue , the service combination solution S _pj joins the queue of the service combination solution S _pi middle.

定义用L_i代表第i层的服务组合解的集合。从第一层(i＝1)开始，初始化当时，找到所有的服务组合解S_pj，并且将该S_pj加入L_i中，然后找到队列中所有的服务组合解从每一个队列中删除服务组合解S_pj，再将服务组合解从RemainingSet中删除。重复上述过程，直至为止，此时可以得到代表解集合一共有q层解。definition Let L _i represent the collection of service composition solutions for the i-th layer. Starting from the first layer (i=1), initialize when , find all The service combination solution S _pj , and add this S _pj to L _i , and then find the queue All service composition solutions in from each queue Delete the service composition solution S _pj in the service composition solution, and then Remove from RemainingSet. Repeat the above process until So far, the set of representative solutions can be obtained at this time There are a total of q layers of solutions.

最后，介绍S600、根据实际服务质量值对所对应的服务组合解进行排序，选择前K个服务组合解作为次优服务组合解集合的步骤。Finally, S600 is introduced, the step of sorting the corresponding service combination solutions according to the actual service quality value, and selecting the top K service combination solutions as the suboptimal service combination solution set.

根据上步骤中实际服务质量值，对所选择的前s层中全部服务组合解进行排序，并选择前K个服务组合解作为次优服务解集合。According to the actual service quality value in the previous step, sort all the service combination solutions in the selected first s layers, and select the top K service combination solutions as the suboptimal service solution set.

与现有技术中考虑服务质量相关性时计算复杂等问题相比，本发明实施例提供的方法能够解决多目标的服务组合问题，可以有效地处理不同服务质量目标之间的权衡及折中；此外，本发明实施例通过计算服务组合解的粗略服务质量进行分层，并计算所选择对应层中候选服务的相关性信息，以获取实际服务质量值，从而能够处理实际系统带有相关性的服务组合情况，能够快速得到实际系统的次优解，从而提高求解效率。Compared with problems such as complex calculation when considering service quality correlation in the prior art, the method provided by the embodiment of the present invention can solve the multi-objective service combination problem, and can effectively deal with the trade-offs and compromises between different service quality goals; In addition, the embodiment of the present invention performs layering by calculating the rough service quality of the service composition solution, and calculates the correlation information of the candidate services in the selected corresponding layer to obtain the actual service quality value, so as to be able to handle the actual system with correlation In the case of service combination, the suboptimal solution of the actual system can be quickly obtained, thereby improving the solution efficiency.

虽然结合附图描述了本发明的实施方式，但是本领域技术人员可以在不脱离本发明的精神和范围的情况下做出各种修改和变型，这样的修改和变型均落入由所附权利要求所限定的范围之内。Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention. within the bounds of the requirements.