CN112306665B - Service integration method driven by sequence QoS - Google Patents

Abstract

The invention discloses a service integration method driven by a sequence QoS, which comprises the following steps: presetting a sequence combination flow chart, and selecting QoS attribute characteristics; processing the QoS attribute characteristics to obtain an initial vector CQoS of a service sequence sc, normalizing the initial vector CQoS to obtain a normalized vector CQoS ', and further calculating a service score ScorCQoS'; extracting a data object subset which is not dominated by any other candidate service object from each candidate service object database corresponding to the task node in the preset sequence combination flow chart by using Skyline calculation, and taking the data object subset as an initial search space for solving an ant colony algorithm; and solving an allocation scheme corresponding to the optimal solution in the initial search space according to the ant colony algorithm. The method solves the problems of data flooding and user preference, and improves the efficiency of the integration process and the service quality of the service sequence.

Description

A sequential QoS-driven service integration method

Technical field

The present invention relates to the technical field of service integration, and in particular to a sequence QoS-driven service integration method.

Background technique

Currently, more and more enterprises, organizations or individuals publish their physical resources to the Internet in the form of services. Service providers virtualize service resources according to business process divisions and build independent service modules. These modules often appear in the form of applications or APIs and can be deployed, run, and managed independently in the environment. Users only need to use lightweight application software, interfaces or APIs (Application Programming Interface) to call services with corresponding functions. However, due to the limited service functions provided by service providers, in many cases a single service may not be able to meet the functional needs of users. In this case, it is necessary to effectively integrate a set of services with limited functions to meet the functional needs of users, and ensure that the combined The service has good quality of service (QoS), but relying on users to solve service integration problems poses certain challenges.

In the real world, services have multiple execution sequences due to different business functions during the integration process, such as sequential execution, concurrent execution, and sequential and concurrent mixed execution. In order to better solve service integration problems, many scholars have given solutions to parallel service serialization. Therefore, only sequentially executed service sequences are considered. As shown in Figure 1, in a service process, each service cluster corresponds to a task node _ti , i∈[1, 5], and each task node _ti corresponds to a candidate service set _Ti , that is, each service A cluster corresponds to a set of candidate services. In order to meet user service needs, a specific service instance should be scheduled from each task node, so that the final composite service composed of several atomic services can provide users with functions that a single atomic service cannot provide.

As shown in Figure 1, since the service instances in the candidate service set are not unique, multiple integration methods are generated, corresponding to multiple service instance sequences. In order to obtain the best service sequence set, this solution also uses service quality as the basis for distinguishing service sequences. However, the difference from atomic services is that the service quality of a service sequence cannot directly consider and directly process the service quality of each atomic service participating in its composition. First of all, the data "flooding" problem and user preference problem caused by the different calculation units of each characteristic attribute of service quality cannot provide a complete solution. Secondly, many integration methods are suitable for small business fields and are not suitable for complex and large-scale services. The efficiency and service quality of the integration process are unsatisfactory.

Contents of the invention

The present invention aims to solve one of the technical problems in the related art, at least to a certain extent.

To this end, one purpose of the present invention is to propose a sequence QoS-driven service integration method, which can improve the solution quality and solution speed of the service integration algorithm.

In order to achieve the above purpose, the embodiment of the present invention proposes a sequence QoS driven service integration method, which includes the following steps: Step S1, preset sequence combination flow chart, select service quality QoS attribute characteristics; Step S2, process the service quality QoS attribute Features, obtain the initial vector CQoS of the service sequence sc, normalize the initial vector CQoS to obtain the standardized vector CQoS', and then calculate the service score ScoreCQoS'; step S3, use Skyline to calculate the preset sequence combination process Extract a subset of data objects that are not dominated by any other candidate service objects from each candidate service object database corresponding to the task node in the figure, and use the subset of data objects as the initial search space for solving the ant colony algorithm; Step S4, according to the The ant colony algorithm solves the allocation plan corresponding to the optimal solution in the initial search space.

The sequence QoS-driven service integration method of the embodiment of the present invention provides a complete definition for the service quality calculation method in business sequence integration, solving the data "flooding" problem in traditional service integration problems and users' differences in different attributes of QoS. The problem of preference; the ant colony algorithm based on Skyline improves the solution quality and solution speed of the service integration algorithm.

In addition, the sequential QoS-driven service integration method according to the above embodiments of the present invention may also have the following additional technical features:

Further, in one embodiment of the present invention, the quality of service QoS attribute characteristics include response time, availability, throughput and success rate.

Further, in one embodiment of the present invention, in step S2, based on the preset sequence combination flow chart, the response time, availability, throughput and success rate in the quality of service QoS attribute characteristics are calculated respectively, Obtain the initial vector CqoS of the service sequence sc.

Optionally, in one embodiment of the present invention, the response time is the sum of the response times of each atomic service call from the start node to the end node; the availability is the time from the start node to the end node. The product of the availability of each atomic service; the throughput is the minimum value of the throughput of each atomic service passed from the start node to the end node; the success rate is the success rate of each atomic service call passed from the start node to the end node Success rate product.

Further, in one embodiment of the present invention, the initial vector CQoS normalization process in step S2 is: specifying the availability, the throughput and the success rate according to the quality of service QoS attribute characteristics is a positive attribute feature, and the response time is specified as a positive and negative attribute feature; perform positive Qos normalization and negative Qos normalization respectively according to the normalization strategy to obtain the standardized vector CQoS'; according to user preference settings and the normalization The vector CQoS' is calculated to obtain the service score ScoreCQoS'.

Further, in one embodiment of the present invention, the ant colony algorithm includes foraging rules, optimal solution calculation rules and pheromone update principles.

Further, in one embodiment of the present invention, the step S4 includes: step S401, selecting a specific service instance for each task node in the initial search space based on the foraging rule; step S402, after all task nodes are assigned service instances, based on the optimal solution calculation rule, using the service score ScoreCQoS’ to measure the quality of the current service combination plan; iteratively executing step S401 until all ants have solved an allocation method and a corresponding service score ScoreCQoS’, that is, the allocation plan corresponding to the optimal solution; step S404, updating pheromones based on the pheromone update principle through the allocation plan corresponding to the optimal solution.

Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of drawings

The above and/or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

Figure 1 is an example diagram of service business sequence;

Figure 2 is a flow chart of a sequential QoS driven service integration method according to an embodiment of the present invention;

Figure 3 is a flow chart of an ant colony algorithm solution according to one embodiment of the present invention;

Figure 4 is a histogram comparison chart of the running time of the ant colony algorithm ACO, Skyline combined with the ant colony algorithm SkylineACO, the reinforcement learning method RL and the improved particle swarm algorithm MPSO according to a specific embodiment of the present invention;

Figure 5 is a schematic diagram of a Skyline calculation time bar chart according to a specific embodiment of the present invention;

Figure 6 is a polyline comparison chart of the solution quality of the ant colony algorithm ACO, Skyline combined with the ant colony algorithm SkylineACO, the reinforcement learning method RL and the improved particle swarm algorithm MPSO according to a specific embodiment of the present invention;

Figure 7 is a comparison chart of the solution quality of Skyline combined with the ant colony algorithm SkylineACO and the reinforcement learning method RL according to a specific embodiment of the present invention;

Figure 8 is a comparison chart of the iteration number sensitivity of the ant colony algorithm ACO, Skyline combined with the ant colony algorithm SkylineACO, the reinforcement learning method RL and the improved particle swarm algorithm MPSO according to a specific embodiment of the present invention, when the number of task nodes is 10;

Figure 9 is a comparison chart of the iteration number sensitivity of the ant colony algorithm ACO, Skyline combined with the ant colony algorithm SkylineACO, the reinforcement learning method RL and the improved particle swarm algorithm MPSO according to a specific embodiment of the present invention, when the number of task nodes is 20.

Detailed ways

Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present invention and are not to be construed as limiting the present invention.

The following describes the sequential QoS-driven service integration method proposed according to the embodiment of the present invention with reference to the accompanying drawings.

Figure 2 is a flow chart of a sequential QoS-driven service integration method according to an embodiment of the present invention.

As shown in Figure 2, this sequential QoS-driven service integration method includes the following steps:

In step S1, a sequence combination flow chart is preset and quality of service QoS attribute characteristics are selected.

It should be noted that the quality of service QoS attribute characteristics selected in the embodiment of the present invention are response time (ResponseTime), availability (Availability), throughput (ThroughPut), and success rate (Successability), but are not limited to these. Technology in the art Personnel can be adjusted according to actual needs.

In step S2, the quality of service QoS attribute characteristics are processed to obtain the initial vector CQoS of the service sequence sc, and the initial vector CQoS is normalized to obtain the standardized vector CQoS', and then the service score ScoreCQoS' is calculated.

Specifically, in order to describe the QoS of the service sequence, the embodiment of the present invention represents the QoS of the atomic service s _ij as a vector QoS (qos _ResponseTime , qos _Availability , qos _ThroughPut , qos _{Successability} ), i represents the task node number, and j represents the task node The candidate service corresponding to i is combined with the internal service instance number. For example, optionally select s ₁₁ ∈T ₁ , s ₂₁ ∈T ₂ , s ₃₁ ∈T ₃ , s ₄₁ ∈T ₄ , s ₅₁ ∈T ₅ , then it consists of s ₁₁ , s ₂₁ , s ₃₁ , s ₄₁ and s ₅₁ One service sequence sc ₁ . The QoS of service sequence sc is expressed as CQoS (cqos _ResponseTime , cqos _Availability , cqos _ThroughPut , cqos _{Successability} ), and its calculation methods are as follows:

Response time: In the preset sequence combination flow chart, the combined service response time is the sum of the response times of each atomic service call from the start node to the end node. Then the response time of the combined service cqos _ResponseTime is calculated by formula (1).

Among them, N represents the number of atomic services involved in the sequential combination service, that is, the number of task nodes, s _ij represents the service instance numbered j in the candidate service set corresponding to the i-th task node, and the size of j depends on the number of instances in the candidate service set. number.

Availability: A preset sequence combination flow chart, whose combined service availability is the product of the availability of each atomic service passing through from the start node to the end node, calculated by formula (2).

Among them, N represents the number of atomic services involved in the sequential combination service, that is, the number of task nodes, s _ij represents the service instance numbered j in the candidate service set corresponding to the i-th task node, and the size of j depends on the number of instances in the candidate service set. number.

Throughput: The preset sequence combination flow chart, the combined service throughput is the minimum value of the throughput of each atomic service passing through from the start node to the end node, calculated by formula (3).

cqos _ThroughPut =Min(qos _ThroughPut (s _ij )),i∈[1,N] (3)

Success rate: The preset sequence combination flow chart has a combination service success rate multiplied by the success rate of each atomic service call from the start node to the end node, calculated by formula (4).

Further, in one embodiment of the present invention, the initial vector CQoS normalization process in step S2 is:

According to the quality of service QoS attribute characteristics, availability, throughput and success rate are specified as positive attribute characteristics, and response time is specified as positive and negative attribute characteristics;

According to the normalization strategy, positive Qos normalization and negative Qos normalization are performed to standardize the initial vector CqoS and obtain the service object set CQoS’.

Specifically, the initial vector CQoS (cqos _ResponseTime , cqos _Availability , cqos _ThroughPut , cqos _{Successability} ) of the service sequence sc can be calculated through the above steps. When calculating the service sequence aggregate value from CQoS, due to the different value ranges of different features, in order to prevent Larger range of data causes smaller range of data to be "flooded", so the CQoS value needs to be standardized and normalized to the range of [0, 1]. According to the quality of service QoS attribute characteristics, availability, throughput, and success rate are positive QoS attribute characteristics, and response time is a negative attribute characteristic. The normalization strategy is divided into positive QoS normalization and negative QoS normalization. The normalization The formulas are as follows: Formula (5) and Formula (6):

front:

Negative:

It should be noted that the cqos _max and cqos _min involved in the normalization process of the CQoS value of the service sequence are the maximum and minimum values among the cqos values of the service sequence corresponding to all possible combination strategies. Take Figure 1 as an example. There are 5 task nodes. Assume that each task node corresponds to 100 candidate services. Then there are a total of 100*100*100*100*100 combination strategies, corresponding to 1005 combination services, resulting in 1005 CQoS value, taking response time as an example, cqos _{ResponseTimeMax} is the maximum response time in CQoS among these 1005 combined services, and cqos _{ResponseTimeMin} is the minimum response time in CQoS among these 1005 service sequences.

Considering that the candidate service set corresponding to each task node may be large, it is impossible to construct all possible combination strategies to calculate all CQoS to find cqos _max and cqos _min . Therefore, the embodiment of the present invention proposes the following calculation method to solve each QoS attribute. cqos _max and cqos _min .

First calculate the qos _max and qos _min of each candidate service set. Still taking the workflow in Figure 1 as an example, there are five candidate service sets T ₁ , T ₂ , T ₃ , T ₄ and T ₅ . Calculate the qos _max and qos _min of each attribute value in the candidate service set respectively. After solving each N maximum values and N minimum values are obtained from the QoS attribute characteristics. N is the number of candidate service sets, which is 5 here. Based on this solution result, cqos _max and cqos _min are calculated. The calculation method of each attribute characteristic is as follows.

Response time: The maximum response time qos _{ResponseTimeMaxi} and the minimum response time qos _{ResponseTimeMini} are calculated based on the candidate service set T _i . Then the maximum response time of the combined service sequence is the sum of the maximum response times of the atomic services in each candidate service set. The combined service sequence The minimum response time is the sum of the minimum response times of the atomic services in each candidate service set. The calculation method of cqos _{ResponseTimeMax} and cqos _{ResponseTimeMin} is as shown in formula (7), where i is the task node number.

Availability: The maximum availability value qos _{AvailabilityMaxi} and the minimum availability value qos _{AvailabilityMini} are calculated based on the candidate service set T _i . Then the maximum availability value of the combined service sequence is the product of the maximum availability of the atomic services in each candidate service set, and the minimum availability value of the combined service sequence is The product of the minimum availability of atomic services in each candidate service set, cqos _{AvailabilityMax} and cqos _{AvailabilityMin} are calculated as shown in formula (8).

Throughput: The maximum throughput value qos _{ThroughPutMaxi} and the minimum throughput value qos _{ThroughPutMini} are calculated based on the candidate service set T _i . Then the maximum throughput of the combined service sequence is the maximum value in qos _{ThroughPutMaxi} , and the minimum throughput of the combined service sequence is qos _{ThroughPutMini} . The minimum value of cqos _{ThroughPutMax} and cqos _{ThroughPutMin} are calculated as shown in formula (9).

Success rate: qos _{SuccessabilityMaxi} and qos _{SuccessabilityMini} are calculated based on the candidate service set T _i . The calculation method of the maximum and minimum success rate of the combined service sequence is similar to that of availability. The calculation of cqos _{SuccessabilityMax} and cqos _{SuccessabilityMin} is as shown in formula (10).

After the above steps, the CQoS (cqos _ResponseTime , cqos _Availability , cqos _ThroughPut , cqos _{Successability} ) of the initial combination service is standardized to CQoS' (cqos _ResponseTime ', cqos _Availability ', cqos _ThroughPut ', cqos _{Successability} '). Considering that users have different preferences for different attributes of QoS, given a QoS feature (qos ₁ , qos ₂ , qos ₃ ,..., qos _m ), set the QoS weight P (p ₁ , p ₂ , p ₃ ,..., p _m ), represents the user's degree of attention to QoS features, and its satisfying constraints are shown in formula (11).

According to the user preference settings and CQOS’, the CQoS’ aggregate value is calculated by formula (12), that is, the service score ScoreCQoS’.

In step S3, Skyline calculation is used to extract a subset of data objects that are not dominated by any other candidate service objects from each candidate service object database corresponding to the task node in the preset sequence combination flow chart, and the data object subset is regarded as an ant colony. The initial search space that the algorithm solves.

It should be noted that skyline computing is to extract a set of data objects from a database that are not dominated by any other data objects. It has potential applications in multi-objective decision-making, data mining, data visualization, etc. A classic example is the hotel selection problem, that is, finding a hotel that is close to the sea and cheap among a large amount of hotel information. In essence, skyline computing is a data extraction method that reflects the internal characteristics of a data set.

For a service object with multi-dimensional QoS information, that is, the QoS attribute characteristics of the service object are not unique, it can be regarded as a data object with multi-dimensional QoS, and the better data object is selected by comparing different data objects, thereby providing Lays the foundation for next service decisions. The purpose of using Skyline calculation in this invention is to extract a subset of data objects that are not dominated by other data objects from the service object set with multi-dimensional service quality, that is, to reduce the service search space and improve the search efficiency of the next step of service combination decision-making.

For example, taking services s1 and s2 as an example, their service qualities are QoS1 (qos ₁ , qos ₂ , qos ₃ , qos ₄ ) and QoS2 (qos ₁ , qos ₂ , qos ₃ , qos ₄ ) respectively. If s ₁ The service quality is better than or equal to s ₂ in all dimensions, and s ₁ is better than s ₂ in at least one dimension, then s ₁ is said to dominate s ₂ .

For positive QoS, its formal definition is as shown in formula (13), where M is the service quality dimension.

For negative QoS, the formal definition is as shown in formula (14).

Through Skyline calculation, service objects that are not dominated by any other candidate service objects are extracted from the candidate service object database corresponding to the task node, thereby forming a Skyline service, which serves as the initial search space for ant colony algorithm solution.

In step S4, the allocation plan corresponding to the optimal solution in the initial search space is solved according to the ant colony algorithm.

It should be noted that the ant colony algorithm is inspired by the foraging behavior of ants in nature. Its basic principle is that ants will release pheromones related to path information on the way to forage for all ants to encounter this location next time. Provide a basis for selection at intersections. During the process of moving forward, if you encounter an intersection that you have not passed, you will choose it randomly. Otherwise, choose the intersection with the highest pheromone concentration. In the end, the pheromone concentration on the optimal path will become larger and larger, and the ant colony will find the optimal path. Excellent foraging paths.

Ant colony algorithm involves the following important processes in solving the optimal service sequence.

Foraging rules: The selection process of ants foraging each time they pass through an intersection is equivalent to selecting a specific service instance from the candidate service set. There are two main types of service instance selection rules. If the current ant number is critical to the ant number of this task node Before the critical point, the maximum pheromone distribution is used, and the ants will select the service instance with the highest pheromone concentration. If it is after the critical point, the service instance will be randomly selected.

For example: taskCritivalPoint[i]=10 means that the critical point of the ant number of the i-th task node is 10. When the ant selects a service instance on the i-th task node, if the ant number is before the 10th (including the 10th), the largest one will be selected. For service instances with pheromone concentration, if the ant number is after No. 10, the service instance will be randomly selected from the candidate service set.

Optimal solution calculation rules: After an ant selects service instances for all task nodes, it obtains an allocation plan, that is, a combined service plan. The standardized combined service score ScoreCQoS' is used to measure the quality of the allocation plan. Standard, where a higher ScoreCQoS' score value represents a better allocation solution.

Pheromone update principle: Pheromones play an important role in ants selecting service instances. For example, phenomoneMatrix[i][j]=0.1 means that the pheromone concentration of selecting the j-th service instance at the i-th task node is 0.1. After each iteration, the pheromone is updated. All pheromone concentrations are attenuated by p%, and the pheromone concentration corresponding to the optimal solution in this iteration is increased by q%.

Therefore, step S4 of the present invention specifically includes:

Step S401: Select a specific service instance for each task node in the initial search space based on foraging rules;

Step S402: After all task nodes are assigned service instances, based on the optimal solution calculation rules, the service score ScoreCQoS’ is used to measure the quality of the current service combination solution;

Step S403, iteratively execute step S401 until all ants have solved the problem, and obtain an allocation method and the corresponding service score ScoreCQoS’, that is, the allocation plan corresponding to the optimal solution;

Step S404: Update the pheromone based on the pheromone update principle through the allocation plan corresponding to the optimal solution.

That is to say, as shown in Figure 3, the combined service process corresponding to the optimal ScoreCQoS' using the ant colony algorithm is: selecting a specific service instance for each task node based on foraging rules, and allocating service instances to all task nodes After completion, the quality of this allocation plan, that is, the service combination plan, is measured based on the optimal solution calculation rules. Repeat the above process until all ants have solved and obtained an allocation method and the corresponding plan quality. The allocation plan corresponding to the optimal solution is based on the information. The pheromone is updated according to the pheromone update principle, and one iteration is completed.

Therefore, the embodiment of the present invention uses Skyline calculation to reduce the large-scale candidate service set and perform local optimization to reduce the solution scale of the ant colony algorithm and improve the solution speed and quality. Then, the ant colony is used according to the QoS of the combined service. The algorithm performs global optimization to find the service instance sequence solution with the best QoS.

A specific embodiment is proposed below to further illustrate the sequential QoS-driven service integration method of the present invention.

This specific embodiment uses an international public data set: the QWS data set to conduct experimental verification of the effectiveness of the algorithm. The QWS data set is a data set generally recognized by researchers in the field of service computing. This specific embodiment retains four QoS attribute information of response time, availability, success rate and throughput in the data set.

In the specific embodiment, workflow templates with 5, 10, 15, 20 and 25 task nodes were designed to simulate different combined service complexities. There were 100 candidate services for each task node, and the Ant Colony Algorithm (ACO), Skyline combines the ant colony algorithm (SkylineACO), reinforcement learning method (RL) and improved particle swarm algorithm (MPSO) to solve the process of combined service solutions. Four tests were conducted in terms of the execution time of the algorithm, the quality of the combined service solution, and the number of iterations. In the comparative experiment of two algorithms, ScoreCQoS' of formula (12) is used as the evaluation standard for the solution quality of the combined service. The higher the score, the better the solution quality.

(1) Algorithm execution time

The number of iterations of ACO, SkylineACO, RL and MPSO is set to 100. The size of the ant colony of ACO, SkylineACO and the particle swarm of MPSO is set to 50. The execution time comparison result of the algorithm is shown in Figure 4. The Skyline calculation time is shown in Figure 5.

Under the same iteration number settings, as the number of task nodes increases, the execution time of the algorithm increases significantly. Among them, the average execution time of the ACO algorithm is significantly higher than other methods. It can be seen that the Skyline ACO algorithm proposed by the present invention can significantly reduce the execution time of the original ACO algorithm. Secondly, the RL algorithm has the lowest average execution time, and SkylineACO is slightly higher than the MPSO algorithm.

(2) Composite service solution quality Based on the experiment in Figure 4, the ScoreCQoS' of the four algorithms under different numbers of task nodes were collected, and multiple experiments were conducted to obtain the average value as the final ScoreCQoS' value of the algorithm. The experimental comparison results are as follows As shown in Figure 6.

It can be seen from Figure 6 that in experiments with different numbers of task nodes, the solution quality of the SkylineACO algorithm proposed by the present invention is optimal, and SkylineACO can significantly improve the solution quality of the original ACO algorithm. In the experimental comparison of 10, 15 and 20 task nodes, the comparison of solution quality is obvious. The quality of SkylineACO solution results is better than the quality of MPSO solution results, and the RL solution quality is the worst. As the number of task nodes increases, the overall solution quality shows a downward trend.

Based on the comparison results in Figure 4 and Figure 6, it can be seen that SkylineACO is superior to the ACO algorithm in terms of algorithm execution time and algorithm solution quality. In terms of algorithm execution time, the RL algorithm has the shortest execution time, and in terms of algorithm solution quality, SkylineACO has the highest solution quality.

For this reason, in order to further compare the SkylineACO and RL algorithms, the execution time of the RL algorithm is increased by increasing the number of iterations of the RL algorithm, so that the execution time of the RL algorithm is roughly the same as that of the SkylineACO algorithm, thereby comparing the solution quality of the two algorithms. The experimental results are shown in Figure 7. The solution quality of the SkylineACO algorithm is better than the solution quality of the RL algorithm.

(3) Iteration number sensitivity

In order to compare the impact of the number of iterations on the solution quality, the experiment collected the solution quality of four algorithms under 50, 100, 150, 200 and 250 iterations. Each algorithm performed 10 experiments under the specified number of iterations to obtain The average combined service score is used as the solution quality criterion. Figures 8 and 9 show the experimental results when the number of task nodes is 10 and the number of task nodes is 20, respectively.

Based on the experimental results in Figures 8 and 9, it is found that the RL algorithm is the most sensitive to the number of iterations. As the number of iterations increases, the solution quality of the RL algorithm improves the most, followed by the SkylineACO algorithm, MPSO has a smaller improvement, and the ACO algorithm has basically no improvement. Inference The ACO algorithm falls into a local optimal solution.

In summary, the experiment proves that using Skyline to calculate the improved ant colony algorithm can significantly improve the execution efficiency of the algorithm and is superior to the comparative method in terms of solution quality. Therefore, the proposed SkylineACO algorithm can quickly find the global optimal Score _CQoS' and its corresponding Combined service plan.

According to the sequence QoS-driven service integration method proposed by the embodiment of the present invention, by giving a complete definition for the service quality calculation method in business sequence integration, the data "flooding" problem in the traditional service integration problem and the different attributes of users for QoS are solved. There are problems with different preferences; the ant colony algorithm based on Skyline improves the solution quality and solution speed of the service integration algorithm.

In addition, the terms “first” and “second” are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of indicated technical features. Therefore, features defined as "first" and "second" may explicitly or implicitly include at least one of these features. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise expressly and specifically limited.

In the description of this specification, reference to the terms "one embodiment," "some embodiments," "an example," "specific examples," or "some examples" or the like means that specific features are described in connection with the embodiment or example. , structures, materials or features are included in at least one embodiment or example of the invention. In this specification, the schematic expressions of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, those skilled in the art may combine and combine different embodiments or examples and features of different embodiments or examples described in this specification unless they are inconsistent with each other.

Although the embodiments of the present invention have been shown and described above, it can be understood that the above-mentioned embodiments are illustrative and should not be construed as limitations of the present invention. Those of ordinary skill in the art can make modifications to the above-mentioned embodiments within the scope of the present invention. The embodiments are subject to changes, modifications, substitutions and variations.

Claims (8)

1. A sequential QoS driven service integration method, comprising the steps of:

step S1, presetting a sequence combination flow chart, and selecting quality of service (QoS) attribute characteristics;

step S2, processing the QoS attribute characteristics to obtain an initial vector CQoS of a service sequence sc, normalizing the initial vector CQoS to obtain a normalized vector CQoS ', and further calculating a service score ScorCQoS';

step S3, using Skyline to calculate and extract a data object subset which is not dominated by any other candidate service object from each candidate service object database corresponding to the task node in the preset sequence combination flow chart, and using the data object subset as an initial search space for solving an ant colony algorithm; and

step S4, solving an allocation scheme corresponding to the optimal solution in the initial search space according to the ant colony algorithm;

wherein the quality of service QoS attribute characteristics include response time, availability, throughput, and success rate;

the initial vector CQoS normalization processing in step S2 is as follows:

according to the QoS attribute characteristics, the availability, the throughput and the success rate are specified as positive attribute characteristics, and the response time is specified as positive and negative attribute characteristics;

respectively carrying out positive Qos normalization and negative Qos normalization according to a normalization strategy to obtain a normalized vector CQoS';

the service score is calculated from the user preference settings and the normalized vector CQoS'.

2. The sequence QoS driven service integration method according to claim 1, wherein in step S2, based on the preset sequence combination flowchart, response time, availability, throughput and success rate in the QoS attribute feature are calculated, respectively, to obtain an initial vector CqoS of the service sequence sc.

3. The sequential QoS-driven service integration method of claim 2, wherein the response time is a sum of response times of atomic service calls passing from a start node to a termination node.

4. The sequential QoS driven service integration method of claim 2, wherein said availability is a product of availability of individual atomic services traversed from a start node to a termination node.

5. The sequential QoS driven service integration method of claim 2, wherein said throughput is a minimum of each atomic service throughput passed from a start node to a termination node.

6. The sequential QoS driven service integration method of claim 2, wherein the success rate is a success rate product of success of each atomic service invocation passed from a start node to a termination node.

7. The sequential QoS driven service integration method of claim 1, wherein the ant colony algorithm comprises foraging rules, optimal solution calculation rules, and pheromone update rules.

8. The sequential QoS-driven service integration method according to claim 7, wherein said step S4 comprises:

step S401, selecting a specific service instance for each task node in the initial search space based on the foraging rule;

step S402, after all task nodes are distributed with service examples, based on the optimal solution calculation rule, the quality of the current service combination scheme is measured by using a service score ScorCQoS';

step S403, iteratively executing step S401 until all ants are solved to obtain an allocation mode and a corresponding service score ScorCQoS', namely an allocation scheme corresponding to the optimal solution;

and step S404, carrying out pheromone updating based on the pheromone updating principle through the allocation scheme corresponding to the optimal solution.