METHOD FOR SYNCHRONIZING LOCAL CLOCKS IN A DISTRIBUTED COMPUTER NETWORK

摘要

The invention relates to a method for synchronizing local clocks in a distributed computer network, where said computer network consists of a number of end systems and at least two switches, where each end system is connected to at least two switches via bidirectional communication links, and where a configured subset of end systems and switches executes the method in form of a synchronization state machine, characterized in that a) the state machine uses at least three different frame types (coldstart frame type, coldstart acknowledge frame type, and integration frame type), and b) the states in the state machine are either said to belong to an unsynchronized set of states or belong to a synchronized set of states, and c) all end systems that are configured as Synchronization Master periodically send coldstart frames in one of the unsynchronized states, and d) all end systems that are configured as Synchronization Master react to the reception of a coldstart frame by sending a coldstart acknowledgment frame a configurable first timeout (CSO) after the reception of the coldstart frame on all replicated communication channels, provided that the end system is in a state in which the synchronization state machine defines a transition for coldstart frames, and where said first timeout (CSO) is reset when a consecutive coldstart frame is received before the coldstart acknowledge is sent, and e) all end systems that are configured as Synchronization Master react to the reception of a coldstart acknowledgment frame by starting a configurable second timeout (CAO), provided that they are not already executing said first timeout (CSO), and entering a synchronized state when said second timeout (CAO) expires.