Re: Did anyone try this replication model?

Nice diagram. Every time I tried something like this in Netscape it spat the dummy at me.

Anyway, on to your question. I think the problem you will hit is in what you've called the roll-up databases. If by that you mean the tables in these databases are summary tables of the main databases (i.e. sums, averages, that sort of thing - anything that will make these snapshots complex rather than simple), they can't be updatable. Complex snapshots by definition must be read-only. If you're not really rolling up the data in some way, you can ignore this and the system will work.

The next thing I'd look at is the need to have 3 masters. Anything more than two masters reduces the options you have for conflict resolution. Without understanding the business needs far more than can be outlined in a newsgroup posting, I can't say whether this will cause you problems, but it definitely needs investigating.

Then, start looking at the network. How much data are you transferring? What are the network speeds? Volume test, volume test, volume test.

HTH. Pete

Atchut wrote:

> [Image] The database, having master-master replication are in
> different sites. Each such database may have roll-up databases in
> subsidiary sites. These Subsidiary Sites have updatable snap shots of
> the tables that are replicated among master-master databases. So any
> change made at a site is reflected in all the other sites in due
> course of time. If any one attempted this model, please tell me what
> care I should take while designing one. What are the possible bottle
> necks and design issues. we are using Oracle 8.0.5 Enterprise
> Servers.

--
Regards

Pete

Received on Fri Jul 23 1999 - 11:12:23 CDT