A practical framework of realizing actuators for autonomous fault management in SOA

Abstract

Due to the key features of Service-oriented architecture (SOA); blackbox-nature of services, heterogeneity, service dynamism, and service evolvability, fault management in SOA is known to be more challenging than conventional system management. An efficient way of managing faults in SOA is to apply principles of autonomic computing (AC), of which process is specified in MAPE. The first two phases of MAPE are to monitor target systems and diagnose faults to determine underlying cause. The other two phases are to plan healing/actuation methods and to execute them. Devising methods to remedy service faults which can run in autonomous manner is a hard problem, mainly due to the remoteness and the limited visibility and controllability. In this paper, we present a practical framework to design actuators which can be invoked autonomously. By considering the relationships among fault, cause, and actuator, we derive the abstract and concrete actuators. For some essential concrete actuators, we present their algorithms which can be implemented in practice. We believe our proposed service actuation framework makes the realization of autonomous service management more feasible. © 2009 IEEE.