JD Edwards E1 9.1.2 - Production vs Non-Production


Hi All,

I am at a client where they have 2 Data Centers and 150 miles separate both the DCs. They are planning to have one location for non-production servers and the other for production servers where the Deployment Server will be located. Based on my understanding of the JD Edwards best practices, it advises to keep all the servers under one roof but I cannot find the document. The servers for the instance are Ent Ser, Trans Ser, BSSV Ser, BIP Ser, Mobility Ser, Exalytics, etc.

I would appreciate your valueable inputs/guidance and any Best Practice document by Oracle would help me back it up with when I raise this issue for advising them to rearchitect.

Thanks very much in advance for your reply.

There's no issue with splitting the solution and we've seen a number of bigger clients using one data centre for non production and disaster recovery with production in the other data centre. Usually they would be running something like Oracle Data Guard which means that the production database is being synchronised between the two data centres. The only thing they need to account for is the bandwidth requirements for a non production full package build but considering the network links that are generally used these days that shouldn't be an issue. 150 miles is nothing in terms of DC separation.

First of all, my recommendation is to try and put everything in one data center. Trying to do what you mention takes time and is complex, and is only right for certain customers. That saying, I have worked with a number of customers in architecting exactly what you state - and can set up the architecture to provide good run-time performance. But the key issue is the location of the deployment server - which services both production and non-production pathcodes. It is possible to put together processes and technical architecture that makes this work - but you need someone who has worked on something like this before ! Give me a call, and I can talk through the high-level areas.
This seems fairly common these days, possibly as auditors like to see a physical separation of production from testing/development. If that is the reason, you may even need to run separate deployment servers for each site anyway. I can't count how many times I've had to explain why developers needs access to a production server.

It doesn't cause any performance problems for production by itself, only in the deployment over WAN, so depends on your connectivity from the deployment server to the remote servers.
Also bare in mind things like, if you refresh test environments from tape backups, do you need to ship the physical tape between data centres. Simple things like this can add a significant cost to the plan.

Recently, I've had:
- completely separated environments including deployments servers in different data centres. Promoting to production could only be done by CNC admin, and involved some manual processes and custom code for some objects and project details (E810). DR was a 3rd data centre.
- separated but with a single deployment server in the production data centre. The dev/test data centre also ran the DR systems. Backups done over the DR box, so easy to restore to dev/test.
- prod hosting dev/test as a VM, DR to a second data centre. Backups done over the DR box so always physically separate from the production data centre.