3 Replies Latest reply on Apr 12, 2008 10:32 AM by Porus Homi Havewala, OCM

    OEM Architecture for Large Organization

      We are implementing OEM Grid Control to monitor a large number of targets (>1000) that are spread out geographically over several data centers.

      I have seen the high availability papers that are available, however I was wondering what best practice would be to create an OEM topology that will handle disparate targets.

      Any papers or advice out there?
        • 1. Re: OEM Architecture for Large Organization
          You can share the same repository. If you want, you could make it an Oracle RAC solution. You might even want to put dataguard in there somewhere if you want failover capabilities but that might be overkill.

          What I would do is setup multiple OMSes for sure. Unfortunately the Agents can only point to one ip so what you could do is use F5 load balancers. By default, OMS and the Agents want to use HTTPS. You can put the certificate on the load balancers if you want (like F5's) or you could just unsecure everything and have the agent go to the load balancer ip and that will put it to a different OMS. The issue you might have though is when you connect to the OMS over a webbrowser, it might try and redirect you to a different URL. That should be fine though since you should be able to directly connect to the OMS anyway.
          • 2. Re: OEM Architecture for Large Organization
            Thanks for the response.

            I am a bit concerned about network latency between data centers. Do you think I should place the load balancer at a centrally located place?

            Also, if I use two geographically seperated OMSes, does it matter which location gets the repository?
            • 3. Re: OEM Architecture for Large Organization
              Porus Homi Havewala, OCM
              Bob, I was the main technical lead for the "EMCentral" project in Telstra Australia
              in 2004 and 2005, which was EM10g Grid Control for all the db servers in the
              company, so I can tell you a few things about the architecture you can use.

              Use commondity servers, Linux and Intel, with 2 cpus each and 8 GB of RAM.

              Have at least 3 or 4 management servers (OMS) all connected to a hardware
              load balancer like Big IP. Have a failover Big IP as well since that is the
              most important thing.

              For the database, you can have it on a single node, or using a standby
              and Oracle Dataguard technology, or you can have a RAC. But the most
              important aspect are the management servers.

              Using such a configuration, you can manage more than a 1000 targets in
              Grid Control.

              In your case, I would suggest putting the database server in 1 central location,
              this is quite important so you have a central view.

              If you are concerned about latency from varous sites, then place management
              servers in your geographical locations. You can have load banancers that
              balance only the nodes in a specific location, That way users can connect
              to the load balancer nearest their area,