A novel Publish/Subscribe System based on Agent Center in e-Science environment

In e-Science environment researchers intend to be informed of events related with their works, including the current states of computing services, service workload and so on. So it is necessary to provide a Publish/Subscribe System in e-Science environment. Also there are some issues in Broker Management in Publish/Subscribe System, such as the abnormal breakdown and the overload of broker. An effective management of brokers can improve the performance of Publish/Subscribe System. In the paper we firstly introduce our Publish/Subscribe System which takes the responsibility for informing researchers of concerned special events. Then we focus on Agent Center, Broker Selection Algorithm based on Standard Deviation and a Balance Algorithm based on Latency. Finally it shows comparisons to describe the characteristics of our system and simulations to evaluate the efficiency of two algorithms.