Time, Clocks and the Ordering of Events in a Distributed System

1
Time, Clocks and the Ordering of Events in a
Distributed System
2
Lamport 1978

• Time, Clocks and the Ordering of Events in a
  Distributed System
• A fully worked out method for synchronizing
  events across a distributed system using the
  concept of logical clock.
• Is the technology required for synchronizing
  sports games.
• Difference between sports games and dead reckoned
  games

3
Lamport 1978

• The first paper that formally details how to
  order events in a distributed system is the
  Lamport,1978 paper entitled Time, clocks, and
  the ordering of events in a distributed system.
• The solution Lamport describes in his paper is
  fairly far reaching it can be used to guarantee
  event ordering in distributed transaction
  systems, networked games, and any other
  networked, distributed collection of processes
  that operate asynchronously.
• The entire notion of how to coordinate
  distributed processes across multiple machines on
  a network asynchronously using a logical clock is
  fundamentally described in this paper.
• A logical clock, as defined by Lamport,1978, is
  just a way of assigning a number to an event,
  where the number is thought of as the time at
  which the event occurred. They may be
  implemented by counters with no actual timing
  mechanism. We more fully detail Lamports
  technology below.

4
(No Transcript)
5
(No Transcript)
6
(No Transcript)
7
Notes Alternatives

• Some issues to think about
• We get events corresponding to a particular tick
  in time and we insert them into the ordered
  queue.
• On a per-client basis, events may be executed in
  a different order BUT systemwise we have assigned
  them to all happening simultaneously.
• Some systems that use variants of this send a
  regular logical clock tick out to all machines on
  the net.
• In fact, they send that tick with UDP and rely on
  the frequent sending to move forward, figuring
  that if a tick is lost, a new one is right behind.

8
Distributed Systems

• In the mid-1980s, distributed transaction
  systems looked over this issue quite frequently.
• Some of the algorithms for reliable transaction
  event ordering are visible in the patent record.

9

• Add time warp material so that we can roll-back
  if we execute too soon.
• Add client/server solution to resolve basketball
  shot/blocked issue roll dice.
• Fix the one pager algorithm so that we save the
  queue after we have executed events so that we
  can send out the last ten time steps to Query
  messages.