CN103209102B - Distributed Measuring System and Method for Web Service Quality - Google Patents

Abstract

The service quality QoS Distributed Measurement System of Web service and a method, system is provided with: Centroid and multiple measured node; It is according to Web service QoS estimation models, utilize network location technology to calculate subscriber's main station, multiple measured node and Web service respectively and dispose main frame, the i.e. service arrangement node network coordinate in abstract network coordinate system, calculate the similarity of subscriber's main station and measured node again, so that the QoS property value that estimating subscriber's can obtain.The present invention can estimate the Web service QoS comprising response time and reliability that user side under different geographical and/or different network environments can obtain quickly and accurately.Method of measurement real-time, does not rely on the history QoS data of user, does not increase added burden to user, and estimation transparent procedures.Innovative point is mapped in multidimensional coordinate system by network system, and in conjunction with the similitude of measured node and subscriber's main station, the response time of the Web service that estimating subscriber's end can obtain and reliability, data are true, credible.<pb pnum="1" />

Description

Distributed Measuring System and Method for Web Service Quality

technical field

The invention relates to a distributed QoS measurement system and method for Web service quality, in particular to a method and system for Web service quality measurement based on a distributed architecture, and belongs to the technical field of computers and network communications.

Background technique

Web service is a service-oriented architecture (Service-Oriented Architecture) technology. It provides services through standard Web protocols to ensure that application services on different platforms can interoperate. Its communication protocol is mainly based on SOAP, and it uses UDDI to discover and obtain services. metadata. Due to the dynamics and unpredictability of the Internet, the rendering results of the same service on different clients are quite different. At the same time, there are many Web services with the same or similar functions in the network. How to choose the appropriate service for users in different regions and different network environments has become a key issue.

In the prior art, there are mainly the following implementation methods related to Web service QoS measurement or estimation:

The document "Distributed QoS evaluation for real-world Web Services" (published in: In Proc.8th IEEE International Conference on Web Services (ICWS2010), 2010,) introduces a distributed QoS automatic measurement method for Web services, and its design purpose is to provide real-world QoS datasets for QoS-based Web service technologies and models. The method measures the reliability, response time and throughput of web services. Firstly, users distributed in different regions call the Web service, and use the measurement results of each user to calculate the overall reliability of the Web service. Second, the actual invocation of the service by the user is used to directly provide data for the QoS set. In this technical solution, users need to participate in the actual Web service invocation process, which increases the user's burden; and the provided data set belongs to the original data, and the overall calculation method of the service QoS attribute does not take into account the characteristics of the service QoS and the region. It is the measurement data of all users, rather than the QoS presented by Web services on the user end, and the result has little practical significance for users.

The document "a scalable hybrid collaborative filtering algorithm for personalized WebService recommendation" (published in: RegionKNN: In Proc.8th IEEE International Conference on Web Services (ICWS2010), 2010) introduces the content: the main body of client QoS estimation technology is based on collaborative filtering algorithm Realized: first calculate the user similarity and Web service similarity, analyze and find out the users who have a certain similarity with the tested user as similar neighbors. Finally, the QoS attribute value of the service that the target client can obtain is estimated by the QoS attribute of similar Web services and the QoS history records of similar neighbors. This paper considers the relevant characteristics of Web service QoS regions, and improves the collaborative filtering algorithm on this basis: firstly, users are segmented according to the geographical location and the similarity of the user's historical QoS data, and then the collaborative filtering algorithm is used to estimate Measure the QoS attribute values that can be obtained by the user end. This method requires the user to provide the system with the QoS data history records obtained by personally calling the Web service. Therefore, the source of its QoS data set is the QoS historical data obtained when users use Web services. Since the QoS attribute value of Web service changes dynamically, relying on the user's QoS historical data will reduce the accuracy of QoS estimation. At the same time, the way to obtain QoS data is unreliable, and the standards obtained are not uniform, so it is difficult to apply it to the actual system.

The system and method proposed in the document "WSP A Network Coordinate based Web Service Positioning Framework for Response Time Prediction." (published in: In Proc.19th IEEE International Conference on Web Services (ICWS2012), 2012) is a distributed framework based on network coordinates , used to estimate the service response time that the client can obtain: use the corresponding relationship between the network coordinate distance and the network delay to calculate the service response time that the client can obtain. The calculation of network coordinates in this method is based on a centralized network coordinate algorithm—GNP algorithm. This method combines the network coordinate positioning technology and the traditional collaborative filtering method to improve the accuracy of response time calculation. The network coordinate algorithm it adopts is based on the GNP method. This centralized coordinate algorithm is inconsistent with the distributed framework adopted by the system. The network delay measured by the distributed deployment nodes must be transmitted to a central node, and the central node completes the network. The calculation of coordinates affects the scalability and the real-time performance of the calculation results.

Contents of the invention

In view of this, the purpose of the present invention is to provide a distributed measurement system and method of Web service quality that can solve the defects of the prior art, and is used to provide users in different regions and/or different network environments quickly and accurately. The QoS attribute value (including response time and reliability) presented by a Web service on the client side; at the same time, the QoS estimation process is transparent to the user, and the user does not need to install any client module. Moreover, the data used for QoS estimation does not depend on user historical data, which can ensure the accuracy of estimation.

In order to achieve the above object, the present invention provides a QoS distributed measurement system of Web service, which is characterized in that: the system calculates the user host, multiple measurement nodes and The coordinates of the Web service deployment host, that is, the service deployment node in the abstract network coordinate system, that is, the network coordinates, and then calculate the similarity between the user host and the measurement node, so as to estimate the QoS attribute value that the user can obtain; the Web service QoS The estimation model is a multi-dimensional coordinate network described in Euclidean space, in which each node including the central node, measurement node, service deployment node and user corresponds to a point in the multi-dimensional coordinate space, and the multi-dimensional coordinate The dimension of the network is the same as the number of measuring nodes; the network coordinate is a coordinate used to predict the distance of the Internet with scalability, which is composed of the current coordinate value of the node and the current error; in the network system, each A node can obtain the network coordinates of the node only after a small amount of measurement; then according to the preset calculation method, using the network coordinates of any two nodes, the network distance between the two can be predicted; the system is equipped with: node and multiple measurement nodes, where:

The central node is responsible for accepting and processing the request of the user host, scheduling the measurement tasks for each measurement node, calculating the similarity between the user host and the measurement node, and analyzing the measurement results; it has four components: task scheduling module and result analyzer , similarity calculation module and database; the functions of the four parts of the central node are as follows:

The task scheduling module, as the control hub of the central node, is responsible for the registration of each measurement node, receiving and responding to the estimation request of the user host, that is, user QoS, and the distribution of the measurement tasks of the entire system: receiving the user including its IP address, When the QoS estimation request of the identified Web service in the system needs to be measured, the task scheduling module sends a measurement command to the measurement node, distributes the user's IP address to all measurement nodes in the system, and assigns measurement tasks; completes the QoS After estimation, return the result to the user;

The result analyzer is used to calculate the service QoS attribute value that the user can obtain including response time and reliability according to the measured network coordinates of the user and the service deployment node, and the similarity between the user and each measurement node, and calculate the result Stored in the database; the response time is that the result analyzer obtains the user coordinates and the service deployment node coordinates from the database, and calculates the network distance between the two, that is, the network transmission of the Simple Object Access Protocol SOAP message between the user and the service deployment host The delay L, plus the processing time of the service, is the response time of the service; the reliability is that the result analyzer calculates each measurement node within the period according to the service call results stored in the database within the set period. The ratio of the number of successful calls to the service to the total number of calls is the reliability value of the service obtained by each measurement node, and then the reliability value of the service that the user can obtain by using the similarity between each measurement node and the user is calculated; The set time period is set by the user;

The similarity calculation module is responsible for obtaining the network coordinates of the user and each measurement node from the database, and calculating the similarity between the user and each measurement node according to the network distance between the user and each measurement node for the task scheduling module and result analysis The device uses the similarity value to select the measurement node and calculate the QoS attribute value that the user can obtain; the similarity between the user and each measurement node in the system depends on the network distance between the two: the closer the distance, the higher the similarity; Conversely, the farther the distance, the lower the similarity;

The database is used to store the latest updated network coordinates of users, measurement nodes and service deployment nodes, as well as the service QoS historical data measured by each measurement node;

Multiple measurement nodes are deployed in the system in a distributed manner, responsible for receiving commands sent by the central node, obtaining and parsing XML language WSDL (Web Services Description Language) files describing Web services and their interface information, and how to communicate with Web services, and performing their own and the positioning of each measurement node, measurement of user host and service deployment node coordinates, and measurement of Web service QoS; there are five in total: measurement node control module, node location update module, WSDL file analysis module, service QoS measurement module and database connection module Components; the five components of this measurement node function as follows:

The measurement node control module is used to receive commands from the central node, communicate interactively with the task scheduling module of the central node, call and control each module of the measurement node to complete various measurement operations;

The node positioning update module is used to update the network coordinates of its own node according to the network transmission delay between the measurement node and other measurement nodes before measuring the Web service QoS; and measure the distance from the measurement node to the service deployment node and the user respectively. Transmission delay to update the network coordinates of the two to ensure the accuracy of the measurement results;

The WSDL file parsing module is used to parse the WSDL file of the web service obtained in the system, and generate the client code required to call the service, and obtain the service name, service operation name, operation parameter name and type, and After the analysis result of the service address is serialized, it is stored in the measurement node for use when measuring the service QoS;

The service QoS measurement module uses the service information obtained by parsing the WSDL file analysis module and the automatically generated client code to obtain the parameter type of the Web service operation and assign the value, and then actually calls the Web service to obtain information including the response time of the service and whether the call is Successful service QoS value;

The database connection module is used to connect to the database for calling by the measurement node control module, node location update module and service QoS measurement module to complete reading, writing, adding and updating of data in the database.

In order to achieve the above object, the present invention also provides a measurement method using the distributed measurement system of the present invention, characterized in that: the method includes the following steps:

Step 1, periodic measurement phase: the WSDL file parsing module of each measurement node parses the IP address of the Web service deployment node, and the node location update module periodically sends ping messages to the IP address and calculates the transmission between its own node and the service deployment node The network delay of the SOAP message; then take the service deployment node as the target node, and use the set node coordinate update method to update the coordinates of the service deployment node; the service QoS measurement module of the measurement node obtains the response time by calling the Web service, and records the call Whether it is successful; in step 1, the measurement node performs the following operations:

(11) The WSDL file parsing module obtains the IP address of the Web service deployment node;

(12) The node positioning update module sends a ping message periodically to the IP address, and records the round-trip time of the message in the network, and then uses the round-trip time as the round-trip transmission time of the SOAP message between the measurement node and the service deployment node;

(13) The node positioning update module takes the service deployment node as the target node, adopts the node coordinate update method set by the system, that is, performs iterative calculations according to the measured network delay between the measurement node and the target node and the current coordinate value of the measurement node, Update the coordinates of the service deployment node; in this step (13), the node coordinate update method adopted by the node location update module includes the following specific operations:

(13a) Calculate the weight of the current coordinates of the target node in the error and calculation Then, from the current coordinates of the target node and the measurement node, the current distance distance between the two is calculated by the method of calculating the distance in Euclidean space; where e and e _n are the current errors of the current coordinates of the target node and the measurement node respectively ;

(13b) Calculate the relative error of the target node coordinates according to the calculated current distance between the two and the measured network distance value And update the coordinate error of the target node accordingly: e=relativeError×Ce×weight+e×(1-Ce×weight), where latency is the measured network delay from the measurement node to the target node, that is, the measured The current distance between the two, Ce is the coefficient that controls the convergence rate during the calculation process;

(13c) Utilize the coordinates of the target node and the measuring node to calculate the moving direction of the target node coordinates, the moving direction should be parallel to the direction of the line connecting the coordinates of the target node and the measuring node, and make the value of e _s decrease; then calculate The moving value ratio of the target node coordinates=Cc×weight×|distance-latency|, where Cc is the convergence parameter of the control algorithm;

(13d) Utilize the calculation result of step (13c) to update the current coordinates of the target node;

(13e) judging whether the error value of the current coordinates of the target node is less than the set threshold, if not, then return to step (13a) to perform the corresponding operation; if so, stop updating the coordinates of the target node;

(14) After the network coordinate update of the service deployment node is completed, the service QoS measurement module obtains the response time by calling the Web service, and records whether the call is successful, and prepares data for estimating the service response time and reliability of the client;

Step 2, real-time measurement phase:

When the central node receives the user's service QoS estimation request, it forwards the request to each measurement node; each measurement node updates its own network coordinates and user coordinates after receiving the request;

The similarity calculation module of the central node calculates the similarity between the user and each measurement node, so that the task scheduling module can select K measurement nodes with the largest similarity with the user;

The K measurement nodes update the user and service deployment node coordinates again, measure the response time of the service and calculate the processing time of the current service deployment node;

The result analyzer of the central node calculates the service response time and reliability that users can obtain;

This step 2 includes the following operations:

(21) The user sends to the central node a Web service QoS estimation request including the user's IP address and the Web service that needs to be measured in the system;

(22) The task scheduling module of the central node distributes the user IP address in the received user request to all measurement nodes in the system;

(23) The node positioning update module of each measurement node updates its own coordinates and user coordinates respectively:

After each measurement node receives the measurement command, under the control of its system control module, it first sends ping messages between each measurement node to perform its own positioning and complete its own coordinate update; all measurement nodes complete their own network coordinate update Afterwards, each measurement node sends a ping message to the user respectively, and measures the round-trip network delay between its own node and the user, and also completes the estimation of the user coordinates according to the node coordinate update method; then The user coordinates are updated to the database;

This step (23) includes the following specific operations:

(23a) The measurement node that currently needs to update its own coordinates sends ping messages to all other measurement nodes in the system through the node positioning update module, and records the round-trip time of the message in the network, and then uses the round-trip time as two measurement nodes The network delay between;

(23b) Obtain the network coordinates of all measuring nodes from the database, and then use these network coordinates and the network time delay obtained in step (23a) as the input data of the node coordinate update algorithm, and then use the current measuring node as the target node, using the The above-mentioned node coordinate updating method calculates and obtains the new network coordinate of each current measuring node;

(23c) updating the original stored network coordinates of the measuring node in the database with the new network coordinates;

(23d) The node positioning update module updates the user's coordinates in the same way as above steps (23a)-(23c);

(24) The similarity calculation module of the central node calculates the similarity:

After all measurement nodes have completed step (23), the similarity calculation module of the central node takes out the updated coordinates of the user and each measurement node from the database, calculates the network distance between the user and each measurement node, and is used for calculating The similarity between the user and each measurement node;

(25) The task scheduling module of the central node selects the measurement node and sends a measurement command: according to the calculated similarity, the task scheduling module selects K nodes with the largest similarity with the user, and sends an order to continue measurement to the K measurement nodes ;

(26) The node positioning update module of the measurement node updates the coordinates of the user and the service deployment node: when the K measurement nodes receive the measurement command, they update the coordinates of the user and the service deployment node respectively, so as to ensure the real-time performance and quality of QoS estimation. accuracy;

(27) The service QoS measurement module of the measurement node measures the response time of the service: according to the analysis service operation name and parameter name and type obtained by the WSDL file analysis module, the service QoS measurement module automatically calls the Web service operation, and records whether the service call is successful, At the same time, record the response time T _r of the measurement node calling the service; then calculate the service processing time T _p according to the formula T _r = T _l + T _p , and store it in the database for use in result analysis; where T _l is The transmission time of the SOAP message between the measurement node and the service deployment node, that is, the network distance calculated from the coordinates of the measurement node and the service deployment node;

(28) The result analyzer of the central node estimates the response time and reliability of the service respectively:

According to the user coordinates in the database and the service deployment node coordinates, the network distance between the two is calculated by the distance calculation method in Euclidean space, that is, the transmission time T _l of the SOAP message between the user and the Web service, Then calculate the response time for the user to obtain the service according to T _r =T _l +T _p ;

According to the service call results stored in the database within the set time period, calculate the reliability P _k of each measurement node k to obtain the service, and the value of this P _k is the success of the measurement node k calling the service within the set time period The ratio of the number of calls to the total number of calls; then using the similarity between the user and the K measurement nodes, according to the formula Calculate the reliability of the service that the user can obtain, where W _k is the proportion of the similarity between the measurement node k and the user in the sum of the similarities between all K measurement nodes and the user;

(29) Return the estimated results, including response time and reliability, to the user.

The key technology of the present invention is: the measuring system and measuring method thereof based on the Web service quality of distributed architecture can quickly and accurately estimate the Web service QoS (including response QoS) that the client can obtain under different regions and/or different network environments. time and reliability), the method of the present invention has strong real-time performance, does not depend on the user's historical QoS data, does not add additional burden to the user, and the estimation process is completely transparent. The innovative point of the method of the present invention is to map the network system into a multi-dimensional coordinate system, combine the similarity between the measurement node and the user host, and estimate the response time and reliability of the Web service presented on the user end, so that the data is true and credible.

Compared with the prior art, the present invention has the advantages of: proposing a QoS estimation method based on the network coordinate model and the similarity between nodes, and the system and method of the present invention use distributed measurement nodes to measure the QoS of Web services based on WSDL. Carry out measurement, and then perform real-time estimation of the service quality presented by the Web service on the user end according to the measurement data, that is, perform QoS estimation on the basis of the actual measurement data. The estimation process is transparent to the user and does not depend on the user's QoS history Data, and the user does not increase any burden, the only requirement for the user is to be able to receive and respond to ping messages, which improves the scalability of the measurement system.

The present invention also considers the characteristics of dynamic change of service QoS, solves the problems of unreliable access to QoS data and inconsistent acquisition standards, and ensures the accuracy of QoS estimation. Furthermore, the network coordinate algorithm adopted in the present invention is a distributed algorithm, the measurement process of each measurement node is independent and does not affect each other, and it is easy to add new measurement nodes; Changes, strong real-time performance, and can meet the needs of users for obtaining real-time quality attributes of services. In addition, the concept of similarity between nodes introduced by the present invention is used to select measurement nodes and calculate results, which saves system overhead and ensures measurement accuracy. Therefore, the present invention has very good application prospects.

Description of drawings

Figure 1 is a schematic diagram of the composition of the distributed measurement system for Web service QoS.

Fig. 2 is a schematic diagram of an estimation model of Web service QoS (the Web service deployment node represents the position of the service deployment host in the geometric space).

FIG. 3 is a flow chart of the operation steps of the method for estimating the QoS of a Web service.

Fig. 4 is a sequence diagram of the stage of real-time estimation of Web service QoS;

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings.

Referring to Fig. 1, the structural composition of Web service QoS distributed measuring system of the present invention is introduced:

According to the Web service QoS estimation model, the system uses network positioning technology to calculate the coordinates of the user host, multiple measurement nodes and Web service deployment hosts (service deployment nodes) in the abstract network coordinate system, and then calculates the coordinates of the user host and Measure the similarity of nodes in order to estimate the QoS attribute value that users can obtain. The system is equipped with: a central node and multiple measurement nodes. in:

The central node is responsible for accepting and processing the request of the user host, scheduling the measurement tasks for each measurement node, calculating the similarity between the user host and the measurement node, and analyzing the measurement results; it has four components: task scheduling module and result analyzer , a similarity calculation module and a database. The function of each component is introduced as follows:

Task scheduling module: As the control hub of the central node, it is responsible for the registration of each measurement node, receiving and responding to the estimation request of the user host, that is, user QoS, and the distribution of measurement tasks for the entire system: receiving users including their IP addresses, When the QoS estimation request of the identified Web service in the system needs to be measured, the task scheduling module sends a measurement command to the measurement node, distributes the user's IP address to all measurement nodes in the system, and assigns measurement tasks; completes the QoS After estimation, the result is returned to the user.

Result Analyzer: It is used to calculate the service QoS attribute values that users can obtain including response time and reliability according to the measured network coordinates of users and service deployment nodes, and the similarity between users and each measurement node, and calculate the results stored in the database;

Similarity calculation module: responsible for obtaining the network coordinates of the user and each measurement node from the database, and calculating the similarity between the user and each measurement node according to the network distance between the user and each measurement node, for the task scheduling module and result analysis The device uses the similarity value to select the measurement node and calculate the QoS attribute value that the user can obtain; the similarity between the user and each measurement node in the system depends on the network distance between the two: the closer the distance, the higher the similarity; Conversely, the farther the distance, the lower the similarity;

Database: used to store the latest updated network coordinates of users, measurement nodes and service deployment nodes, and service QoS historical data measured by each measurement node.

Multiple measurement nodes are deployed in the system in a distributed manner, responsible for receiving commands sent by the central node, obtaining and parsing XML language WSDL (Web Services Description Language) files describing Web services and their interface information, and how to communicate with Web services, and performing their own and the positioning of each measurement node, measurement of user host and service deployment node coordinates, and measurement of Web service QoS; there are five in total: measurement node control module, node location update module, WSDL file analysis module, service QoS measurement module and database connection module part. The function description of each part is as follows:

Measurement node control module: used to receive commands from the central node, interact with the task scheduling module of the central node, call and control each module of the measurement node to complete various measurement operations.

Node positioning update module: used to update the network coordinates of its own node according to the network transmission delay between the measurement node and other measurement nodes before measuring the Web service QoS; and measure the distance from the measurement node to the service deployment node and the user respectively Transmission delay to update the network coordinates of the two to ensure the accuracy of the measurement results.

WSDL file parsing module: used to parse the WSDL file of the web service obtained in the system, and generate the client code needed to call the service, and obtain the service name, service operation name, operation parameter name and type, and After the analysis result of the service address is serialized, it is stored in the measurement node for use when measuring the service QoS.

Service QoS measurement module: Use the service information obtained by parsing the WSDL file analysis module and the automatically generated client code to obtain the parameter type of the Web service operation and assign the value, and then make an actual call to the Web service to obtain information including the response time of the service and whether the call is Successful service QoS value.

Database connection module: used to connect to the database for calling by the measurement node control module, node location update module and service QoS measurement module to complete reading, writing, adding and updating of data in the database.

The network coordinate of the system of the present invention is a coordinate used for predicting the Internet distance with scalability, and the network coordinate is composed of the current coordinate value of the node and the current error. In the network system, each node only needs a small amount of measurement to obtain the network coordinates of the node; then according to the preset calculation method, using the network coordinates of any two nodes, the network distance between the two can be predicted.

Referring to FIG. 2 , a brief introduction of the multi-dimensional coordinate network described in Euclidean space (shown as a three-dimensional coordinate system) selected by the Web service QoS estimation model of the system of the present invention is introduced. The whole system is equipped with: central node, measurement node, service deployment node and user, where each node corresponds to a coordinate point in the multi-dimensional coordinate space, and the dimension of the multi-dimensional coordinate network is the same as the number of measurement nodes. Number all the N measurement nodes in the system one by one in sequence, and set the initial value of the k-th dimension coordinate of the k-th measurement node to 0, and the initial values of the coordinates of other dimensions are random values, and each coordinate value Both are dynamically updated as the network status changes (wherein, both N and k are natural numbers, and k is less than or equal to N). The network coordinate system established as above can calculate the similarity between the user and each measurement node through the different coordinates of the target user node, service deployment node, and measurement node, and calculate the Web service QoS attribute value (user The most concerned about is the service response time and reliability) to estimate, the present invention is used for measuring and estimating these two Web service QoS attribute values, briefly explain these two parameters below:

(A) Response time: It is the most important indicator to measure service performance, and it has the characteristics of being geographically related and end-to-end. Web service calls need to transmit SOAP messages through the network to complete the transmission of service requests and responses, and the delivery time of SOAP messages is related to network delay; at the same time, service response time is also related to the load of the server. Therefore, the response time of service invocation is decomposed: response time T _r =T _l +T _p , where T _l is the network round-trip transmission time for the user to invoke the service to deliver the SOAP message, and T _p is the processing execution time of the service. That is to say, the response time is that the result analyzer obtains the user coordinates and the service deployment node coordinates from the database, and calculates the network distance between the two, that is, the simple object access protocol SOAP (Simple Object Access Protocol) message between the user and the service deployment host. The sum of the network transmission delay T _l between and the processing time T _p of this service.

The analysis found that the process of invoking services between the measurement node and the user host is exactly the same, that is to say, there is similarity between them. When the distance between the user and the measurement node is closer, the similarity is higher; the farther the two are, the similarity the lower the degree. First, all the N measurement nodes in the system are respectively numbered as {1,2,...,N}, by calculating a measurement node a(a∈{1,2,...,N}) to other measurement nodes The network distance L _a,i of node i (i∈{1,2,...,N} and i≠a, that is, the remaining measurement nodes except a) and the network distance L _{u,i of user u to measurement node i} , the similarity between measurement node a and user u can be calculated. For example, the formula for calculating similarity using Pearson correlation coefficient: Calculate the similarity between the two; where Sim(u,a) is the similarity between node a and user u, is the mean value of Lu _u,i , is the average value of L _a,i . It should be noted that the formula for calculating the similarity by using the network distance is not limited to the Pearson correlation coefficient, which is just an example here.

Similarity plays a very important role in user's Web service QoS estimation. The system of the present invention sorts all measurement nodes according to the calculated similarity, and selects K nodes with the highest similarity with users from all measurement nodes to update the coordinates of users and service deployment nodes. This not only saves the consumption of system resources, but also ensures the accuracy of measured user and service deployment node coordinates.

According to the updated user coordinates and service deployment node coordinates, the network distance between the two is calculated by the distance calculation method in Euclidean space, that is, the transmission time T _l of the SOAP message between the user and the Web service, Then calculate the response time for the user to obtain the service according to T _r =T _l +T _p .

(B) Reliability: The reliability of a Web service is the probability that a user successfully invokes a Web service. Since users in different network environments and regions may obtain different service reliability, the reliability measured by a certain measurement node or the overall reliability obtained by all measurement nodes cannot represent the service reliability that users can obtain. The system of the present invention still uses the similarity between the user and the service measurement node to estimate the reliability.

Reliability is the ratio of the number of times each measurement node calls the service to the total number of times the service is called by the result analyzer according to the service call results stored in the database within the set period of time, which is the number of times each measurement node obtains the service. Reliability value, and then use the similarity between each measurement node and the user to calculate the reliability value that the user can obtain the service.

Assuming that the reliability of the service obtained by the measurement node k is P _k , the calculation formula for the reliability P of the service obtained by the user is: In the formula, W _k is the proportion of the similarity between the measurement node k and the user in the sum of the similarities between all K measurement nodes and the user.

Referring to Fig. 3, two operation steps of the process flow of the method for estimating QoS in the distributed measurement system of the present invention are introduced:

Step 1, periodic measurement phase: the WSDL file parsing module of each measurement node parses the IP address of the Web service deployment node, and the node location update module periodically sends ping messages to the IP address and calculates the transmission between its own node and the service deployment node The network time delay of SOAP message; then take the service deployment node as the target node, and update the coordinates of the service deployment node by using the set node coordinate update method (the coordinate update algorithm is not only used to update the network coordinates of the service deployment node, but also used for Update the network coordinates of a user and a measurement node); the service QoS measurement module of the measurement node obtains the response time by calling the Web service, and records whether the call is successful.

In step 1, the measurement node performs the following specific operations:

(11) The WSDL file parsing module obtains the IP address of the Web service deployment node.

(12) The node location update module periodically sends a ping message to the IP address, and records the round-trip time of the message in the network, and then uses the round-trip time as the round-trip transmission time of the SOAP message between the measurement node and the service deployment node.

(13) The node positioning update module takes the service deployment node as the target node, adopts the node coordinate update method set by the system, that is, performs iterative calculations according to the measured network delay between the measurement node and the target node and the current coordinate value of the measurement node, Update the coordinates of the service deployment node.

Wherein, the node coordinate update method adopted by the node location update module includes the following specific contents:

(13a) Calculate the weight of the current coordinates of the target node in the error and calculation Then, from the current coordinates of the target node and the measurement node, the current distance distance between the two is calculated by the method of calculating the distance in Euclidean space; where e and e _n are the current errors of the current coordinates of the target node and the measurement node respectively .

(13b) Calculate the relative error of the target node coordinates according to the calculated current distance between the two and the measured network distance value And update the coordinate error of the target node accordingly: e=relativeError×Ce×weight+e×(1-Ce×weight), where latency is the measured network delay from the measurement node to the target node, that is, the measured The current distance between the two, Ce is the coefficient that controls the convergence rate during the calculation.

(13c) Utilize the coordinates of the target node and the measuring node to calculate the moving direction of the target node coordinates, the moving direction should be parallel to the direction of the line connecting the coordinates of the target node and the measuring node, and make the value of e _s decrease; then calculate The moving value ratio of the target node coordinates=Cc×weight×|distance-latency|, where Cc is the convergence parameter of the control algorithm.

(13d) Update the current coordinates of the target node by using the calculation result of step (13c).

(13e) Judging whether the error value of the current coordinates of the target node is less than the set threshold, if not, then return to step (13a) to perform the corresponding operation; if so, stop updating the coordinates of the target node.

(14) After the update of the network coordinates of the service deployment node is completed, the service QoS measurement module obtains the response time by calling the Web service, and records whether the call is successful, and prepares data for estimating the service response time and reliability of the client.

Step 2, real-time measurement phase:

When the central node receives the user's service QoS estimation request, it forwards the request to each measurement node; each measurement node updates its own network coordinates and user coordinates after receiving the request;

The similarity calculation module of the central node calculates the similarity between the user and each measurement node, so that the task scheduling module can select K measurement nodes with the largest similarity with the user;

The K measurement nodes update the user and service deployment node coordinates again, measure the response time of the service and calculate the processing time of the current service deployment node;

The result analyzer of the central node calculates the service response time and reliability that the user can obtain.

See Figure 4 to introduce the specific operation content of step 2:

(21) The user sends to the central node a Web service QoS estimation request including the IP address of the user and the identification of the Web service to be measured in the system.

(22) The task scheduling module of the central node distributes the user IP address in the received user request to all the measurement nodes in the system.

(23) The node positioning update module of each measurement node updates its own coordinates and user coordinates respectively:

After each measurement node receives the measurement command, under the control of its system control module, it first sends ping messages between each measurement node to perform its own positioning and complete its own coordinate update; all measurement nodes complete their own network coordinate update Afterwards, each measurement node sends a ping message to the user respectively, and measures the round-trip network delay between its own node and the user, and also completes the estimation of the user coordinates according to the node coordinate update method; then The user coordinates are updated to the database. This step includes the following specific operations:

(23a) The measurement node that currently needs to update its own coordinates sends ping messages to all other measurement nodes in the system through the node positioning update module, and records the round-trip time of the message in the network, and then uses the round-trip time as two measurement nodes The network delay between;

(23b) Obtain the network coordinates of all measurement nodes from the database, and then use these network coordinates and the network time delay obtained in step (23a) as the input data of the node coordinate update algorithm, and then, with the current measurement node as the target node, use node The coordinate update method calculates the new network coordinates of each current measurement node;

(23c) updating the original stored network coordinates of the measuring node in the database with the new network coordinates;

(23d) The node location update module updates the user's coordinates in the same way as the above steps (23a)-(23c).

(24) The similarity calculation module of the central node calculates the similarity:

After all measurement nodes have completed step (23), the similarity calculation module of the central node takes out the updated coordinates of the user and each measurement node from the database, calculates the network distance between the user and each measurement node, and is used for calculating The similarity between the user and each measurement node;

(25) The task scheduling module of the central node selects the measurement node and sends a measurement command: according to the calculated similarity, the task scheduling module selects K nodes with the largest similarity with the user, and sends an order to continue measurement to the K measurement nodes ;

(26) The node positioning update module of the measurement node updates the coordinates of the user and the service deployment node: when the K measurement nodes receive the measurement command, they update the coordinates of the user and the service deployment node respectively, so as to ensure the real-time performance and quality of QoS estimation. accuracy;

(27) The service QoS measurement module of the measurement node measures the response time of the service: according to the analysis service operation name and parameter name and type obtained by the WSDL file analysis module, the service QoS measurement module automatically calls the Web service operation, and records whether the service call is successful, At the same time, record the response time T _r of the measurement node calling the service; then calculate the service processing time T _p according to the formula T _r = T _l + T _p , and store it in the database for use in result analysis; where T _l is The transmission time of the SOAP message between the measurement node and the service deployment node, that is, the network distance calculated from the coordinates of the measurement node and the service deployment node;

(28) The result analyzer of the central node estimates the response time and reliability of the service respectively:

According to the user coordinates in the database and the service deployment node coordinates, the network distance between the two is calculated by the distance calculation method in Euclidean space, that is, the transmission time T _l of the SOAP message between the user and the Web service, Then calculate the response time for the user to obtain the service according to T _r =T _l +T _p ;

According to the service call results stored in the database within the set time period, calculate the reliability P _k of each measurement node k to obtain the service, and the value of this P _k is the success of the measurement node k calling the service within the set time period The ratio of the number of calls to the total number of calls; then using the similarity between the user and the K measurement nodes, according to the formula Calculate the reliability of the service that the user can obtain, where W _k is the proportion of the similarity between the measurement node k and the user in the sum of the similarities between all K measurement nodes and the user;

(29) Return the estimated results, including response time and reliability, to the user.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the present invention. within the scope of protection.

Claims (3)

1. A quality of service QoS distributed measurement system of Web service, it is characterized in that: this system utilizes network location technology to calculate respectively user host computer, a plurality of measurement nodes and Web service deployment host computer, namely service according to Web service QoS estimation model Deploy the coordinates of the nodes in the abstract network coordinate system, that is, the network coordinates, and then calculate the similarity between the user host and the measurement node, so as to estimate the QoS attribute value that the user can obtain; the Web service QoS estimation model is based on Euclidean The multi-dimensional coordinate network described by Reed space, each node including the central node, measurement node, service deployment node and user corresponds to a point in the multi-dimensional coordinate space, and the dimension of the multi-dimensional coordinate network is the same as that of the measurement node The number is the same; the network coordinate is a coordinate used to predict the distance of the Internet with scalability, which is composed of the current coordinate value of the node and the current error; in the network system, each node only needs a small amount of measurement The network coordinates of the node can be obtained; then according to the preset calculation method, using the network coordinates of any two nodes, the network distance between the two can be predicted; the system is equipped with: a central node and multiple measurement nodes, in: The central node is responsible for accepting and processing the request of the user host, scheduling the measurement tasks for each measurement node, calculating the similarity between the user host and the measurement node, and analyzing the measurement results; it has four components: task scheduling module and result analyzer , similarity calculation module and database; the functions of the four parts of the central node are as follows: The task scheduling module, as the control hub of the central node, is responsible for the registration of each measurement node, receiving and responding to the estimation request of the user host, that is, user QoS, and the distribution of the measurement tasks of the entire system: receiving the user including its IP address, When the QoS estimation request of the identified Web service in the system needs to be measured, the task scheduling module sends a measurement command to the measurement node, distributes the user's IP address to all measurement nodes in the system, and assigns measurement tasks; completes the QoS After estimation, return the result to the user; The result analyzer is used to calculate the service QoS attribute value that the user can obtain including response time and reliability according to the measured network coordinates of the user and the service deployment node, and the similarity between the user and each measurement node, and calculate the result Stored in the database; the response time is that the result analyzer obtains the user coordinates and the service deployment node coordinates from the database, and calculates the network distance between the two, that is, the network transmission of the Simple Object Access Protocol SOAP message between the user and the service deployment host The delay L, plus the processing time of the service, is the response time of the service; the reliability is that the result analyzer calculates each measurement node within the period according to the service call results stored in the database within the set period. The ratio of the number of successful calls to the service to the total number of calls is the reliability value of the service obtained by each measurement node, and then the reliability value of the service that the user can obtain by using the similarity between each measurement node and the user is calculated; The set time period is set by the user; The similarity calculation module is responsible for obtaining the network coordinates of the user and each measurement node from the database, and calculating the similarity between the user and each measurement node according to the network distance between the user and each measurement node for the task scheduling module and result analysis The device uses the similarity value to select the measurement node and calculate the QoS attribute value that the user can obtain; the similarity between the user and each measurement node in the system depends on the network distance between the two: the closer the distance, the higher the similarity; Conversely, the farther the distance, the lower the similarity; The database is used to store the latest updated network coordinates of users, measurement nodes and service deployment nodes, as well as the service QoS historical data measured by each measurement node; Multiple measurement nodes are deployed in the system in a distributed manner, responsible for receiving commands sent by the central node, obtaining and parsing XML language WSDL files describing Web services and their interface information, and how to communicate with Web services, and positioning themselves and each measurement node , measure user host and service deployment node coordinates, and measure Web service QoS; there are five components: a measurement node control module, a node location update module, a WSDL file analysis module, a service QoS measurement module, and a database connection module; the measurement node The functions of the five components are as follows: The measurement node control module is used to receive commands from the central node, communicate interactively with the task scheduling module of the central node, call and control each module of the measurement node to complete various measurement operations; The node positioning update module is used to update the network coordinates of its own node according to the network transmission delay between the measurement node and other measurement nodes before measuring the Web service QoS; and measure the distance from the measurement node to the service deployment node and the user respectively. Transmission delay to update the network coordinates of the two to ensure the accuracy of the measurement results; The WSDL file parsing module is used to parse the WSDL file of the web service obtained in the system, and generate the client code required to call the service, and obtain the service name, service operation name, operation parameter name and type, and After the analysis result of the service address is serialized, it is stored in the measurement node for use when measuring the service QoS; The service QoS measurement module uses the service information obtained by parsing the WSDL file analysis module and the automatically generated client code to obtain the parameter type of the Web service operation and assign the value, and then actually calls the Web service to obtain information including the response time of the service and whether the call is Successful service QoS value; The database connection module is used to connect to the database for calling by the measurement node control module, node location update module and service QoS measurement module to complete reading, writing, adding and updating of data in the database. 2. The distributed measurement system according to claim 1, characterized in that: in the Web service QoS estimation model, all N measurement nodes in the system are numbered one by one in sequence, and the kth The initial value of the k-th dimension coordinate of a measurement node is 0, and the initial values of the other dimension coordinates are random values, and each coordinate value is dynamically updated as the network status changes. The network coordinate system thus established can be obtained by The different coordinates of the target user node, service deployment node, and measurement node, calculate the similarity between the user and each measurement node, and estimate the service QoS attribute value that the target user can obtain, including service response time and reliability, where N and k are both natural numbers, and k is less than or equal to N. 3. A measurement method of the distributed measurement system according to claim 1, characterized in that: the method comprises the following steps: Step 1, periodic measurement phase: the WSDL file parsing module of each measurement node parses the IP address of the Web service deployment node, and the node location update module periodically sends ping messages to the IP address and calculates the transmission between its own node and the service deployment node The network delay of the SOAP message; then take the service deployment node as the target node, and use the set node coordinate update method to update the coordinates of the service deployment node; the service QoS measurement module of the measurement node obtains the response time by calling the Web service, and records the call Whether it is successful; in step 1, the measurement node performs the following operations: (11) The WSDL file parsing module obtains the IP address of the Web service deployment node; (12) The node positioning update module sends a ping message periodically to the IP address, and records the round-trip time of the message in the network, and then uses the round-trip time as the round-trip transmission time of the SOAP message between the measurement node and the service deployment node; (13) The node positioning update module takes the service deployment node as the target node, adopts the node coordinate update method set by the system, that is, performs iterative calculations according to the measured network delay between the measurement node and the target node and the current coordinate value of the measurement node, Update the coordinates of the service deployment node; in this step (13), the node coordinate update method adopted by the node location update module includes the following specific operations: (13a) Calculate the weight of the current coordinates of the target node in the error and calculation Then, from the current coordinates of the target node and the measurement node, the current distance distance between the two is calculated by the method of calculating the distance in Euclidean space; where e and e _n are the current errors of the current coordinates of the target node and the measurement node respectively ; (13b) Calculate the relative error of the target node coordinates according to the calculated current distance between the two and the measured network distance value And update the coordinate error of the target node accordingly: e=relativeError×Ce×weight+e×(1-Ce×weight), where latency is the measured network delay from the measurement node to the target node, that is, the measured The current distance between the two, Ce is the coefficient that controls the convergence rate during the calculation process; (13c) Utilize the coordinates of the target node and the measuring node to calculate the moving direction of the target node coordinates, the moving direction should be parallel to the direction of the line connecting the coordinates of the target node and the measuring node, and make the value of e _s decrease; then calculate The moving value ratio of the target node coordinates=Cc×weight×|distance-latency|, where Cc is the convergence parameter of the control algorithm; (13d) Utilize the calculation result of step (13c) to update the current coordinates of the target node; (13e) judging whether the error value of the current coordinates of the target node is less than the set threshold, if not, then return to step (13a) to perform the corresponding operation; if so, stop updating the coordinates of the target node; (14) After the network coordinate update of the service deployment node is completed, the service QoS measurement module obtains the response time by calling the Web service, and records whether the call is successful, and prepares data for estimating the service response time and reliability of the client; Step 2, real-time measurement phase: When the central node receives the user's service QoS estimation request, it forwards the request to each measurement node; each measurement node updates its own network coordinates and user coordinates after receiving the request; The similarity calculation module of the central node calculates the similarity between the user and each measurement node, so that the task scheduling module can select K measurement nodes with the largest similarity with the user; The K measurement nodes update the user and service deployment node coordinates again, measure the response time of the service and calculate the processing time of the current service deployment node; The result analyzer of the central node calculates the service response time and reliability that users can obtain; This step 2 includes the following operations: (21) The user sends to the central node a Web service QoS estimation request including the user's IP address and the Web service that needs to be measured in the system; (22) The task scheduling module of the central node distributes the user IP address in the received user request to all measurement nodes in the system; (23) The node positioning update module of each measurement node updates its own coordinates and user coordinates respectively: After each measurement node receives the measurement command, under the control of its system control module, it first sends ping messages between each measurement node to perform its own positioning and complete its own coordinate update; all measurement nodes complete their own network coordinate update Afterwards, each measurement node sends a ping message to the user respectively, and measures the round-trip network delay between its own node and the user, and also completes the estimation of the user coordinates according to the node coordinate update method; then The user coordinates are updated to the database; This step (23) comprises following specific operation steps: (23a) The measurement node that currently needs to update its own coordinates sends ping messages to all other measurement nodes in the system through the node positioning update module, and records the round-trip time of the message in the network, and then uses the round-trip time as two measurement nodes The network delay between; (23b) Obtain the network coordinates of all measuring nodes from the database, and then use these network coordinates and the network time delay obtained in step (23a) as the input data of the node coordinate update algorithm, and then use the current measuring node as the target node, using the The above-mentioned node coordinate updating method calculates and obtains the new network coordinate of each current measurement node; (23c) updating the original stored network coordinates of the measuring node in the database with the new network coordinates; (23d) The node positioning update module updates the user's coordinates in the same way as above steps (23a)-(23c); (24) The similarity calculation module of the central node calculates the similarity: After all measurement nodes have completed step (23), the similarity calculation module of the central node takes out the updated coordinates of the user and each measurement node from the database, calculates the network distance between the user and each measurement node, and is used for calculating The similarity between the user and each measurement node; (25) The task scheduling module of the central node selects the measurement node and sends a measurement command: according to the calculated similarity, the task scheduling module selects K nodes with the largest similarity with the user, and sends an order to continue measurement to the K measurement nodes ; (26) The node positioning update module of the measurement node updates the coordinates of the user and the service deployment node: when the K measurement nodes receive the measurement command, they update the coordinates of the user and the service deployment node respectively, so as to ensure the real-time performance and quality of QoS estimation. accuracy; (27) The service QoS measurement module of the measurement node measures the response time of the service: according to the analysis service operation name and parameter name and type obtained by the WSDL file analysis module, the service QoS measurement module automatically calls the Web service operation, and records whether the service call is successful, At the same time, record the response time T _r of the measurement node calling the service; then calculate the service processing time T _p according to the formula T _r = T _l + T _p , and store it in the database for use in result analysis; where T _l is The transmission time of the SOAP message between the measurement node and the service deployment node, that is, the network distance calculated from the coordinates of the measurement node and the service deployment node; (28) The result analyzer of the central node estimates the response time and reliability of the service respectively: According to the user coordinates in the database and the service deployment node coordinates, the network distance between the two is calculated by the distance calculation method in Euclidean space, that is, the transmission time T _l of the SOAP message between the user and the Web service, Then calculate the response time for the user to obtain the service according to T _r =T _l +T _p ; According to the service call results stored in the database within the set time period, calculate the reliability P _k of each measurement node k to obtain the service, and the value of this P _k is the success of the measurement node k calling the service within the set time period The ratio of the number of calls to the total number of calls; then using the similarity between the user and the K measurement nodes, according to the formula Calculate the reliability of the service that the user can obtain, where W _k is the proportion of the similarity between the measurement node k and the user in the sum of the similarities between all K measurement nodes and the user; (29) Return the estimated results, including response time and reliability, to the user.