Michael Boman wrote:
> A server goes down [and the mail should been taken care of by
> another server, automatically and samlessly.]
>
> A single point of failure is not an option.
>
> Best regards
> Michael Boman
At the cost of more WAN traffic, you could add patches so
that on delivery failures, in addition to a message being added to
the local queue it also gets copied to one or more other peers for
queuing. Whenever a message that was queued gets successfully
delivered, a notification message is sent to the associated peer,
so it can dequeue its redundant message.
Implementing this would require:
full description of the redundancy protocol
implementing the protocol in the software
Depending on the various costs (WAN bandwitdh, CPU, storage space,
programmer time) this architecture could result in something similar
to usenet, with each extended queue storage server contacting the
others at regular intervals with a list of message-IDs it has received,
so that all of them get multiple chances to receive the same message
stuck in the queue.
--
David Nicol 816.235.1187 [EMAIL PROTECTED]
drawn to the speed and performance
