New tutorial

Joe Armstrong <>
Wed Feb 5 14:58:00 CET 2003

Hello everybody,

  A  while  back Todd  Proebsting  asked me  how  you  build a  simple
fault-tolerant server.

  After a moment's thought I said "that's easy"

  After two moments thought I said "that's difficult"

  and after three moments thought I said "It's possible".

  To make something fault-tolerant you  need at least two machines and
you need  to replicate things -  if you think  you can do this  on one
machine stop reading now and go play with visual basic ...

  I made  myself a  little client-server. The  server is  actually two
servers and and  data on these two servers  is replicated using shared
mnesia disk tables.

  The client knows about both servers - if server one is down it tries
server two and vice versa, this is similar to DNS - the alternative is
some complicated IP-failover stuff - which is very non-portable.

  Transactions between the client and (virtual) server are idempotent.

  The solution  I came up  with is not  entirely obvious (at  least it
wasn't to me) but it is (hopefully) simple enough to be grokked if you
know a small amount of Erlang.

  The solution and all the code are available at:

  All comments, corrections and improvements are welcomed.


More information about the erlang-questions mailing list