Research On The Key Technologies Of Distributed Web Services Selection With QoS Guarantee

With the development and popularization of Internet technology,more and more web services are appearing in the network.In such an open service network with a large number of services,services can be widely distributed in different physical locations,meanwhile subordinated to different business organizations.How to manage these distributed services effectively,how to access these distributed services efficiently by query,and how to call these distributed services flexibly by composition has becomes an urgent problem.QoS refers to non-functional attributes of web services,such as availability,reliability,response time,reputation and so on.When a user invokes a service request,it is usually required for non-functional attributes,in addition to the requirements of the functional attributes.At present,most of the service selection methods based on QoS are centralized.According to the technical path of the centralized algorithm,it selects the optimal service or service composition on a central node that log global service information.However,it is clear that such methods are no longer applicable to the current widely distributed service architecture.Based on the analysis of the current situation of service selection and the development trend of distributed service architecture,this dissertation studies on the four problems of distributed service architecture,single distributed service selection,distributed combination service selection and service ranking.The specific contents are as follows:1.We present a distributed UDDIs architecture with active monitoring mechanism.Although there are already some researches on distributed service architecture,they mostly use the mixed distributed structure.This structure means that there is a lightweight central node in the overall system architecture.The node does not register the web service,but it is responsible for maintaining the information of each distributed registration node.This architecture improves the robustness of the system by reducing the stress of the central node,but the single-point-of-failure still not to be ignored.With the continuous expansion of the scale of web services applications,the centralized service registry UDDI in the traditional SOA architecture will eventually become the bottleneck of system development.In this dissertation,a distributed multi-UDDI architecture with active monitoring mechanism is proposed.These distributed service UDDIs nodes constitute web service overlay in the form of P2P network to realize distributed service registration,query selection and combination of call and other functions.Moreover,in this architecture to establish an active monitoring mechanism.It can effectively monitor the running status of distributed service nodes in real time,and improve the real-time validity of web services information.2.We propose a distributed service selection algorithm that can consider the global service state.At present,most of the researches on service selection algorithms assume that web services are centrally registered.All service attribute information and execution logs are stored in a central node.However,in a real service-computing environment,a functional requirement may have a large number of web services that meet the constraints,and these services are widely distributed on multiple different service nodes on the Internet.These physically dispersed services are interconnected over the network,and if centralized service selection algorithms are used directly,all service information that meets functional constraints needs to be transmitted over the network,which leads to unacceptable communication overhead and query delay.Based on the skyline query algorithm,this paper makes a cross-node query on the distributed service space,and obtains the distributed global selection of web services with relatively small network traffic consumption.The service selection method can not only achieve gradual output,but also flexibly and dynamically adjust according to user's subjective preferences.3.We present a dynamic distributed composition service selection algorithm based on execution planning.In the study of the existing service compositon methods,there is always a centralized service registry center.The QoS attribute information for all alternative web services will exist on it.The service selection method can take into account the topology of the composite service and the global optimal web service in any case.However,in a distributed service environment,each service registration node only understands a portion of the composite service topology and the QoS attributes of the associated alternative services.Therefore,this paper proposes a service selection method,which is applicable to the distributed architecture and autonomy of each service registration node to optimize the overall combined service selection process.The algorithm performs the selection of sequential iterations according to the combinatorial service logic.Finally,a series of experiments under the simulated distributed service architecture environment verify the effectiveness of the proposed algorithm.4.A Web service ranking method that considering QoS attribute index equilibrium constraints is proposed.In the process of Web service selection,service ranking is an indispensable part.Ranking is to set a set of Web services according to certain.rules into an ordered sequence.Service ranking can not only initialize and filter the service set in the initial stage of service selection to improve the selection efficiency,but also produce service selection results and then further optimize the service set.In this paper,the service ranking method based on QoS attributes is converted to a multi-criteria decision-making problem.Firstly,a subjective and objective combination weight computing method is proposed,and then a QoS index expectation function is constructed according to the multi-dimensional indexes in service QoS attributes.On this basis,it calculates the QoS value of each service,the expectation distance of the balance and the utility value of the service,and then ranks the output.In the experimental part,an example is used to describe the ranking process.Moreover,the effectiveness and flexibility of the method is proved by multiple iterative experiments on real data.