7 Replies Latest reply: Apr 23, 2012 12:08 PM by 911519 RSS

    Infrastructure required for EPM 11.1.2.x in a Distributed Environment

    911519
      Hi All,

      Brief overview of what I'm trying to do:
      - I'm have downloaded EPM 11.1.2.1 components from eDelivery and plan to go in for a component-wise installation.
      - Aim: I would like to understand the required underlying infrastructure and the possible alternatives for a distributed environment..

      Current infrastructure plan:
      a. 2xFoundation UNIX servers set-up as an active-active cluster with load-balancing and (automatic) failover. Still working on how the automatic failover would work.. Any inputs on this would be great :)
      b. 2xEssbase UNIX servers set-up as Active-passive with (automatic/manual) failover. For this excercise, I can't go for active-active as it would require the DB to be read-only and thus won't support DB write-back which is generally required.

      As per the Installation guide > Deploying to a Shared Drive Environment (UNIX):
      1. Map all the machines in the deployment to the same network share.
      2. During installation, on the Destination/Middleware Home page, specify a directory on the shared drive.
      -->
      So here we are not using the local server storage for all the servers. But we install each in a directory in the shared drive.
      Say a common NAS shared drive was mounted at /opt/shared/common_drive/ on all the 4 servers.
      Then we would install:
      Foundation server-1 on /opt/shared/common_drive/installation/FndSrv_1
      Essbase server-1 on /opt/shared/common_drive/installation/EssSrv_1
      ...and so on.
      3. For each machine, on the Configure a New or Existing EPM Oracle Instance configuration page, for Home directory for EPM Oracle instances specify the same instance home, and for EPM Oracle Instance name specify a new instance name.
      --> Say Home directory for EPM Oracle instances is at: /opt/shared/common_drive/EPM_Oracle_Instances_Home
      And for each Server, I specify a new instance name like EPM_FndSrv1, EPM_EssSrv1, ...

      Before I go further, I would like your inputs on this.. Are the paths valid?
      Also am I correct in assuming that the foundation servers need to have access to each others /../user_projects (EPM_Oracle_Instances_Home path) for load-balancing and failover to work?

      Kindly share your views on this..
      Thanks in advance
        • 1. Re: Infrastructure required for EPM 11.1.2.x in a Distributed Environment
          854045
          You should start with the EPM's Standard Deployment Guide. It should answer most of your questions.

          http://www.oracle.com/technetwork/middleware/bi-foundation/epm-deploy-guide11121-1401625.pdf
          • 2. Re: Infrastructure required for EPM 11.1.2.x in a Distributed Environment
            911519
            Hi Kash Mohammadi,

            Thanks for your inputs..

            Yes I have gone through the deployment guide as well. As per the notes - I have come up with my understanding of the sample directory structure. But as per the examples shared in the guides, the EPM Oracle Instance home directory appears to be on a dedicated disk and not shared across the servers..
            So that leaves my questions unanswered. Pls let me know if I have missed out something though.. :)

            Edited by: 908516 on Apr 17, 2012 11:05 AM
            • 3. Re: Infrastructure required for EPM 11.1.2.x in a Distributed Environment
              854045
              Did you look at the doc I sent you. It's not the installation guide but a steps by step guide for the whole deployment process. Just follow it as it's the best practice unless there is some specific customer requirement that is forcing you not to use it.
              • 4. Re: Infrastructure required for EPM 11.1.2.x in a Distributed Environment
                Pablo
                Automatic failover works along with a load balancer like Big-IP's F5. You create a pool for two (or more) OHS servers. Then each OHS server has two or more welogic servers. For Shared Services you need a cifs share for the import_export folder and one for the Reporting and Analysis filesystem store. Also, if you use FDM you will have to create a share for the two FDM servers. That is obviously the 10,000 feet view. You can't specify the same instance in a different active member. Also, Unix works if a) you don't plan to use HFM for reporting and b) if you don't use EPMA. There is a lot of detail and not an question that can be simply answered in a short forum response.

                hth,

                Pablo
                • 5. Re: Infrastructure required for EPM 11.1.2.x in a Distributed Environment
                  JohnGoodwin
                  The standard deployment guide is based on windows 2008 r2 and I think the questions being raised are about the installation on unix.
                  As far as I am deploying to a shared drive on unix is one of the options and that is not forced, deplying to a shared location gives advantages such as only having to perform the installation once, each server configuration would have their own unique instance name.

                  Cheers

                  John
                  http://john-goodwin.blogspot.com/
                  • 6. Re: Infrastructure required for EPM 11.1.2.x in a Distributed Environment
                    911519
                    Hi Pablo,

                    Thanks for your inputs

                    - I am a bit familiar with F5's BIG/IP load-balancing methods - round-robin, least connections mode and dynamic ratio - while intelligently supporting session persistence.
                    - We can also manage load balancing via the WebLogic Admin console, and as you have noted by the OHS as well - which I am not familiar with...

                    This is a newbie question - wouldn't having 3 different agents managing load-balancing complicate things..? As the WebLogic server sits on top of the OHS, I guess they work together to provide load-balancing and configuring the WebLogic for clustering/load-balancing should affect the OHS configuration as well. Is this how it works at the high-level or is it more complicated?

                    The EPM System Configurator creates the required cluster and adds servers to the cluster when we deploy the Web applications in the final step of the configuration. So we need not manually configure WebLogic for clustering. But when and where does one configure load-balancing..?

                    Thanks again.. Essbase infrastructure is indeed a vast topic as it is interesting... :)
                    • 7. Re: Infrastructure required for EPM 11.1.2.x in a Distributed Environment
                      911519
                      Hi John,

                      Thanks!
                      I wasn't sure if there were some internal s/w components that required shared access to certain directories for (probably) load-balancing or fail-over.

                      Cheers