|
|
A master time daemon measures the time differences between the clock of the host on which it is running and those of all the other hosts on the network. The master computes the network time as the average of the times provided by slaves on hosts whose clocks are not faulty. In this scheme, a host's clock is considered faulty if its time differs from that of the majority of the clocks on the network by more than a specified amount. The master time daemon then sends to each slave time daemon the correction that should be applied to its host's clock. This process is repeated periodically.
Because the correction is expressed as a time difference rather than an absolute time, transmission delays do not interfere with the accuracy of the synchronization. Whenever a host comes up on the network, it starts a slave time daemon that asks the master for the correct time and resets the host's clock before any user activity can begin. The time daemons are thus able to maintain a single network time in spite of the drift of clocks away from each other. The present implementation of timed is capable of keeping processor clocks synchronized to within 20 milliseconds, but some clocks, because of their design, cannot be adjusted to within an interval smaller than 1 second.