Research On Technology Of Selection Of Services Based On QoS Constraints Of Clients

Along with the development of Internet,Service-Oriented Architecture (SOA) is widely adopted in various applications.Traditional service selection on functionality cannot satisfy requirements of clients.As a result,Quality of Service(QoS) has been the key point of service selecting process.There are two main challenges for service selection with QoS constraint,one of which is how to improve selecting speed and the other is how to improve selecting accuracy. For most of existing service selecting models with QoS constraint,selecting speed is largely limited by network latency,meanwhile the ignorance of individuality of clients impacts selecting accuracy.This paper is supported by 863 Program under Grant No.2007AA01Z139, Research on service description,selection and evaluation based on closed-loop feedback mechanism under Internet environment.Focusing on heterogeneous environments and individuality of clients,this paper proposes similarity conception among clients,which originated from similarity theory in data mining.Similar clients are aware of similar QoS for a same service.This paper designs an extension of Euclidean Distance and puts forward a new algorithm to calculate similarity among clients based on QoS historical records of services. With calculated similarity and selected QoS historical records of services,this paper also proposes a QoS predicting algorithm,which takes full account of individuality of clients and remarkably improves accuracy of service selection.In order to speed up service selection and take advantage of historical service records,this paper builds a service selecting model of multiple-level cache architecture which derived from computer architecture.This model uses similarities among clients to group clients.Each group is mapped to a Cache which deals with service requests and holds the service invocation information of the group.The Caches also use QoS predicting algorithm to select service which results in minimizing the impact of network latency.This paper implements a prototype of service selecting model with multiple-level Cache architecture.A great deal of experiment is executed to test and evaluate the accuracy and speed of this prototype.As while,this paper analyzes the performance of this model and summarizes the key factors effecting accuracy and selecting speed.The result of experiment has shown that this model is effective and efficient and provides a new mechanism for service selection with QoS constraints of clients.