Research And Implementation On A Distributed Service Registry Based On HADOOP Platform

As the SOA (Service Oriented Architecture) and service computing are becoming popular in the IT industry, searching and discovering the web services precisely are drawing more and more attention. The main research method is to extend the descriptive information of web services, for example, by adding semantic information or QoS (Quality of Service) information. It makes the service documents expand continuously. In addition, the threshold of becoming a service provider is lowered, which results in the explosion of web services. The traditional relational database can meet the complex inquiry, but it's incompetent when dealing with service data sets in large scale. The massive number of services has brought many challenges to the traditional service registries. We should improve data storage and process capability of service registry to answer the challenge.The Cloud composes of hundreds of thousands servers machines, so it's competent to scale down and up. It can offer huge storage capability and computing capability. As the cloud computing is becoming more and more popular and mature, it is possible for us to store and compute massive service data by making good use of the advantages of cloud computing. This paper proposes a model of Hadoop-based Distributed Service Registry (HDSR), which has the capabilities of both behavior-aware and QoS-aware service matching via analyzing the Web service registry and cloud computing platform. This thesis studies the implementation based on Hadoop platform of the distributed storage and parallel query of service data. In HDSR, HDFS (Hadoop Distributed File System) and HBase store service data and MapReduce framework processes data. HDSR has three sub-modules, which are Query Parse Module, Behavior-aware Matchmaker Module, and QoS-aware Matchmaker Module. A service request firstly is delivered to the Query Parse Module, and then is parsed into two sub-requests which are functional request and non-functional request. These two requests are dispatched to Behavior-aware Matchmaker Module and QoS-aware Matchmaker Module respectively. The Behavior-aware Matchmaker Module and the QoS-aware Matchmaker Module are the two core modules of the system. This thesis presents a performance evaluation of HDSR. The results demonstrate the feasibility and efficiency of HDSR. At last, this thesis presents some problems to the current research work and the research job to be solved in the future.Since the categories and number of web services are increasing rapidly, the implementation of service data distributed storage and parallel query based on Hadoop platform has very important significance. This thesis gives a meaningful reference for designing a distributed service registry.