Re: location transparency / replication

From: Selwyn C Wies <scwies_at_world.std.com>
Date: Fri, 16 Dec 1994 05:26:12 GMT
Message-ID: <D0w1rp.Fy3_at_world.std.com>


willyk_at_kbigate.stgt.sub.org (Willy Klotz) writes:

>If I have a distributed database (across 2 or more nodes) and want to
>implement location transparency (so that the application does not
>know where the tables actually are stored), I can do this defining
>synonyms pointing to the different oracle instances.
 

>Is it possible to use replicated tables in this environment ? And to
>do inserts / updates on the tables without modification of the
>application ?
 

>What I want to do is something like this:
 

>Database A Database B
 

>Table_ONE ---replicated to ---> TABLE_ONE
 

> ^ I
> I reads
> I I
> I V
> +-------< --- updates ------ Application

>If this is _not_ possible, what other options do I have ?
 

>Willy Klotz
 

>======================================================================
>Willys Mail FidoNet 2:2474/117 2:2474/118 Mail Only System
> CIS: 100020,3517 USR Courier HST dual standard
> willyk_at_kbigate.stgt.sub.org
> -> No Request from 06.00 to 08.00 <-
>======================================================================

Do you need the 2 databases to be in sync immediately? Are the same set of fields updated on both sides ? You can try use database triggers on each of the tables(with DB Links between the Databases) .

Use a flag to stop endless firing of these triggers.  (see Applications Developers Guide - chapter on Triggers) The disadvantage of this is that if either node is down, updates to both sides will be stopped. Your application will get passed a SQL Error message if the remote update fails.

Another option may be Snapshots, although I think bi-directional snapshots are only supported in the latest V7 release.

I hope I understood the ? correctly

scwies_at_word.std.com Received on Fri Dec 16 1994 - 06:26:12 CET

Original text of this message