A Reconfigurable High Availability Infrastructure in Cluster for Grid

The paper presents the implementation and analysis of a servicebased reconfigurable High Availability infrastructure of cluster system for Grid. Based on service notion, the High Availability infrastructure is constructed for mission critical applications on Grid. The application high availability service is responsible for registered applications’ high availability. The Service Manager is in charge of monitoring all services via heartbeat and automatically recovering from failure to assure service high availability. To improve flexibility and dependability, a dynamic reconfiguration mechanism is designed so that the configurations can be changed without interrupting application’s running. A dependability modeling and simulation tool is deployed in the reconfiguration procedure. Dependability parameters can be inputted into the simulation tool. With the objects and sub-model library in the tool, corresponding global model will be constructed and statistic result for different recovery strategies is given. The configuration of application can be adjusted based on these statistic results.