Active-Passive Clustering is supported even in 5.3. Officially I think MSCS was the documented route, and this is documented in the CS 5.3 Installation guide. When I was with a customer we set up (non-standard) AP clustering with IBM HA, though I can’t speak officially to the support protocol on that one.

Active-Active Clustering is supported by setting up multiple Content Servers on the same repository. The idea is then to load balance the CSes by setting up a DocBroker for each Content Server. The docbroker.ini should be configured to talk to a primary Content Server, and then you should add the remaining content servers as backup docbrokers. You can also set the proximity value to tell the docbroker the sequence of CSes it should visit in the case of failover.

Then you have two options:
Option 1 – put a load balancer in front of the docbrokers, and point all app servers to the load balancer.
Option 2 – assuming the app servers are already load balanced, for each app server/client, in the dfc.properties, you want to configure a primary docbroker and then secondaries as backups. This will give you a similar effect to load balancing if the app servers are already load balanced (but not the exact same effect).

This was possible in 5.3 and still possible today.

Thanks,
Ahson

Of course it can! OpenMigrate does migrations at the API layer so moving to 7.0 or 7.1 will not be difficult especially if you are sticking with Webtop as the front end. A few of our clients are planning to do the same thing – upgrade the backend to 7.1 eventually, keeping with Webtop 6.7. The next major move (D2, HPI, etc.) is a big decision and will take some significant analysis and prototyping.

Ellen

I don’t know if 6.7 SP2 is going to be the last release. They just keep saying that it is not being enhanced. Your assumption may be correct, but I cannot confirm that based on what was said at EMC World.

Ellen

We didn’t hear anything new in regards to clustering around the database server. Will check other sources but our thoughts are that clustering in the content server is not available OOTB.

Dave

Ellen

Thanks for your outstanding coverage of EMC World. I really enjoy catching up on these blog posts to keep my finger on the pulse at EMC.

I did have one question which I didn’t see addressed in your blogs from EMC World, and that is around how well high-availability is built into the Content Server. You included a lot of highlights about the number of documents and/or migrations per hour or day, but key to many users is to make certain they can keep that architecture running 24×7. While most companies now can cluster/load balance many of the components of Documentum (web servers, rendering engines, publication tools), the challenge has always been at the heart – the Content Server. My experiences through 5.3 was that only federation was available to host mulitple content servers, and my research into early versions of 6.x was that clustering was not out-of-the-box and required professional services. Did you hear if EMC has made any improvements in the 6.7 or 7.0 platforms to make clustering content servers OOTB?

Just curious – with all the talk you heard about the change in the object model in v7.0, is any of it tied to the suggestion of year’s ago (I think EMC World 2010 or 2011) that EMC would introduce an object model stored a kind of “partial” record so that redundant or empty data wouldn’t take up as much room in the database? I was always worried about the sound of that. It seemed too much like a solution looking for a problem that had more risk than benefit. Do people really complain that their database is too big or takes too much time when we have technology such as clustered/RAC databases and 64-bit addressable RAM in the database server?