One of the scenarios is to simply start sequence with increasing (+1)
value for each system and use increment by (total count of systems +
some overhead for future systems).
So for example you have 3 dbs, the sequences for each system would be as follows
create sequence seq start with 1 increment by 10;
create sequence seq start with 2 increment by 10;
create sequence seq start with 3 increment by 10;

Gints Plivna
http://www.gplivna.eu


2007/7/18, Paula Stankus <paulastankus@(protected)>:
> Guys,
>
> We are working with 8i and 10g databases and a plan to migration to 10g. We
> have a number of various systems that must generate a unique identifier
> across many different databases. Instead of using one sequence with
> database links, we think it is best to eliminate a single-point-of-failure
> and to avoid performance issues by breaking up the sequence into ranges.
> The range would be assigned to a database instance and multiple related
> applications housed in that one database instance would use that range of
> sequence numbers. Our concerns with this approach are running out of
> sequence numbers faster. Eventually all of these systems feed a mainframe
> system that requires a unique transaction i.d. for financial processing.
> That transaction field on the mainframe IMS databases can not be larger than
> 9 digits because there is a moratorium on development on the mainframe
> system.
>
> Does anyone have advice on handling a sequence across multiple databases for
> an entire enterprise?
>
> Thanks in advance.
> Paula
>
> ________________________________
> Sick sense of humor? Visit Yahoo! TV's Comedy with an Edge to see what's on,
> when.
>
>
--
http://www.freelists.org/webpage/oracle-l