I was wondering about how widespread synchronous replication is, and also
what is the alternative solution. I don't have a good grip of how critical
conflicts can be resolved with asynchronous replication, when asynchronous
replication is used. For example, in a financial transaction, I and
somebody else (on my behalf )could overdraw my account by near-simultaneous
withdrawals in different geographic locations. How would asynchronous
replication work in this case?
"Howard J. Rogers" <howardjr2000_at_yahoo.com.au> wrote in message
news:pan.2003.02.13.18.14.34.197295_at_yahoo.com.au...
> On Wed, 12 Feb 2003 22:41:59 +0000, Ralf Fernan wrote:
>
> > Does using synchronous multimaster replication imply that every network
> > link between N master sites has to be up, otherwise (if any of N-1 links
> > is down) a transaction at a master site will fail? So in other words,
> > such a system is actually less available than a non-replicated database?
>
> It certainly does, which is why synchronous replication is very rarely
> implemented. Otherwise, a commit's not a commit until it's been applied to
> every master site. If one site's down, it's a rollback.
>
> Regards
> HJR
Received on Thu Feb 13 2003 - 17:23:46 CST
