Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Mailing Lists -> Oracle-L -> RE: multi-master replication

RE: multi-master replication

From: Steve Ollig <>
Date: Thu, 20 Oct 2005 14:24:43 -0500
Message-ID: <DFA54221DF26D911BC5100B0D079D97A072ABDA3@EXCHMN5>

Greg -  

I've had no experience with Oracle replication. But I did work on a project that did "multi-master replication" with a large number of transactions in a short amount of time. That ambiguous enough for everyone? :o)  

That application had a large central database that contained all data and over 200 "remote" databases each with a subset of central. The remote subsets could overlap (i.e. same row in more than 1 remote database). Transactions (inserts, updates, and deletes) could originate from any of the remotes or Central. All transactions flowed from remote to central and then back out to any remote that also subscribed to the row. Or if the transaction originated at Central it simply replicated to subscribing remotes. This app certainly was not data center agnostic.  

The technology used was Sybase replication. I believe that technology was/is similar to streams. With Sybase replication there were bottlenecks so throughput was sometimes an issue. Though we were able to work around most of the performance issues.  

IMHO the bigger problem was the "chaos" you refer to. It's an extremely complex architecture and very difficult to build and maintain. My question would be what is the business problem you are trying to solve. To quote a guy i read about named Larry - there's more than one way to do it. I usually opt for the simpler solution. Being data center agnostic isn't the root business problem. There are many other ways to achieve HA.  

FWIW - it's been a few years since I worked on that app. But I hear it is still in use though significantly simplified. But the base architecture remains intact. This is the app's busy season and I hear it's holding up well. Though historically that has not always been the case.

-----Original Message-----
From: []On Behalf Of Loughmiller, Greg
Sent: Thursday, October 20, 2005 9:57 AM To: ''
Subject: multi-master replication

hey folks,  

I believe I may have sent this out before - so go ahead and hammer me accordingly:-) But I have an interesting requirement for an effort......  

Has anyone had any experience in trying to do multi-master replication to keep multiple databases in sync? The assumption is that there would be multiple application/database infrastructures - and the data needs to be consistent between the *stacks*. And in one specific scenario - there could be several million txn's per a 12 hour window of the day:-). A couple of the goals and objectives for this chaos are to try to be data center agnostic for the application-thus the need for the data to be consistent. The additional assumption is that this would provide a level of HA in the event there is a problem with one or more application stack (hmmm..... smells like RAC, taste like RAC - maybe it is RAC but not data center agnostic). The version of Oracle is 9206(for now), on Solaris, using HDS/EMC storage.....  

So I guess my questions are:

  1. Can streams keep up in this type of transactional volume?
  2. I have my own opinion of this matter - but have you all seen other methods to maintain the data in such a fashion?
  3. Other technologies for this type of problem?
  4. Or better yet - other potential solutions (don't do multi master but do X,Y,Z )

My personal opinion with the limited exposure to Streams is that it probably could not maintain that type of throughput, as well as placing additional stress on the host as well as the DB. So I thought would throw this out there to get some opinions, thrashing, or ideas from the list.....  

Thanks in advance

Greg Loughmiller  

This email has been scanned by the MessageLabs Email Security System. For more information please visit
Received on Thu Oct 20 2005 - 14:40:41 CDT

Original text of this message