초록

용어보기논문에서 용어와 풀이말을 자동 추출한 결과로, 시범 서비스 중입니다. AI-Helper

Abstract ▼ AI-Helper

The web service usage rate is increasing due to propagation of network. Accordingly, service-oriented architecture (SOA), composition of several independent web services, is getting the spotlight because of its characteristics such as loosely coupled architecture, fast delivery, cost-effectiveness a...

The web service usage rate is increasing due to propagation of network. Accordingly, service-oriented architecture (SOA), composition of several independent web services, is getting the spotlight because of its characteristics such as loosely coupled architecture, fast delivery, cost-effectiveness and so on. For these reasons, more and more enterprises have started to adopt SOA. To build a reliable SOA system, obtaining reliable atomic web service is essential. However, because the invocation environment of each service is different and the web service is provided through network, different service user may obtain different performance. It brings un-expectable failure of system and profit loss. So, conventional research proposed web service reliability prediction methods. Although, previous approach can predict reliability well, the prediction accuracy decreases when there is not enough pre-recorded reliability information. Also, some approaches require addition information to predict reliability. It makes reliability prediction relies on information provided by service provider. In this paper, we define user/service sensitivity if the user or service is on unique environment and has small number of similar users or services. Then user or service is classified as sensitivity or non-sensitivity. Reliability is predicted according to its sensitivity classification. This approach overcome lack of information problem and doesn't require additional information. To evaluate prediction accuracy, we analysed data set consisted of 150 users and 100 services. According to experiment, our approach increased prediction accuracy most 81%, at least 26%. Also, our approach shows higher prediction accuracy in case of composite service.

주제어