Re: Distributed Databases and Two-Phase Commit

In article <2r0cvu$m09_at_charm.magnus.acs.ohio-state.edu>, Patrick S Hart <phart_at_magnus.acs.ohio-state.edu> wrote:
>I am querying this NG for a possible solution to a problem. I am running a
>distributed system with a very slow transmission rate. I would like to update
>the remote machine in intervals as oppossed to a two-phase commit. I was
>wondering if there is a way to determine all the changes to one database during
>a time frame, and with that info., update the other server?
>
>Thanks for your support,
>P.
You may wanna think twice before doing this unless it is not a critical application.Two phase commit guarantees consistency against failures and prserves the integrity of data which would not be the case in ur solution. what if the server crashes in the interval u don't update the database? what if someone queries the data present in all the servers in the time u don't update the database.the solution of delayed update is too optimistic. Two phase commit on the other hand is too pessimistic but it works well. I guess u could use PRO*C to exchange signals between the machines and keep a record of the transactions and some savepoints and set-up a timeframe to update the databases.
Just my $2/100 worth of opinion.

Thanks and Regards
Mahesh Vallampati
M.S. In EE
Dept.of Electrical Engineering,
Texas A & M University.
Ph:(409)845-6189
\\ In the Beginning there was Codd..... Received on Fri May 13 1994 - 20:35:53 CEST