A Service Discovery System Analyzed with a Queueing Theory Model

Service discovery is indispensable to both resource sharing and service integration in the Internet/Grid. In the face of enormous and various services which scatter in the Internet/Grid, the central issue needed to be addressed by a service discovery system is how to quickly shrink the search scope and exactly locate the desired services. Moreover, since simulation experiments are limited in the number of server nodes, the number and diversification of services etc., which are much less than the real ones in the Internet/Grid, it is necessary to predict the behaviors of service discovery systems and check whether these systems reach their design goals before deploying them by theoretical modeling. This paper develops a service discovery system named SCN4I, which builds an accurate index in the form of a distributed suffix tree to speed up the discovery procedure. Moreover, the paper employs open Jackson network to model SCN4I. Experiments are also conducted to verify the model. Furthermore the model is used to calculate SCN4I's performance bounds and help choose proper deployment policies. Our modeling approach can also be applied to the other service discovery systems for the Internet/Grid.