Synctime

1. SYNCHRONIZING PHYSICAL TIME: central server
2. time server and distributed to nodes. Latency between transfer could be a problem.
   • 估计来回的时间
   • 注意不要回到过去
   • 可以miss tick放慢时间
3. 没有一个标准时间的话，就设立一个master询问每个client时间，average.调整时间是原来的两倍。
4. BERKELEY ALGORITHM: server向所有人请求时间，算入RTT，average，返回client，注意drift可能是原来的两倍。
5. logical time: 保证顺序。高时间不可能导致低时间的任务发生。
6. LAMPORT LOGICAL TIME: happened before
   • The counter is incremented before each event.
   • The message should carry the new (incremented) timestamp.
   • 维护全局index，不能处理接受和发送的问题
   • If two events have no message exchange, Lamport logical time cannot tell which one happened first
   • 可能出现无法处理因果的情况
7. google truetime: compares itself with others, twice drift
8. Cristian's Algorithm is one approach to synchronizing physical clocks using a time server.