Abstract: In order to avoid the intrinsic drawbacks of the centralized fault management in a computer network or a communication network, an alarm correlation algorithm using distributed intelligent agents in a SNMP network management environment is proposed in this paper. The algorithm can be divided into three steps. （1） The intelligent agent resides on a network device. Once it receives an alarm emitted by the device, it classifies the alarm according to its meaning and its severity level, and then inputs the result to a finite state machine （FSM） of the fault model of a certain protocol object. Next, it runs the FSM in order to establish the current state of the protocol object. （2） Using the formal definition of a correlation relation, the intelligent agent tries to correlate the protocol objects whose current states are abnormal. （3） In order to add a reliability measurement to the correlation result, the intelligent agent computes a belief of the correlation result protocol object, and then reports the object and its belief to a network manager. This algorithm decomposes the complexity of the alarm correlation and fault identification, so it is of great benefit to realize real time and automated network management.