|
There are two major requirements that affect the answer:
What is the desired application availability? If you have global, 24X7 production, you will want a distributed system, probably featuring RAC (Real Application Clusters), which at one time was called parallel server. If you are unsure of the availability you need, ask yourself or your operations management personnel what the cost of the system being down at any time. Then ask your financial management what the dollar cost would be if you could not process orders/shipments/payments for a period of time. Can the users tolerate an outage over night? For a couple of days? How adept are your DBAs at database/application troubleshooting?
What is the cost of lost data? This will help drive your backup and recovery plans. If your system fails, and an hour (or days) transactions are lost, what is the cost for manual recovery/rekeying data and transactions lost? Whatever the availability required is, you will want a carefully planned (and tested) backup and recovery plan.
These questions are separate from the production/development/test instance question. You always want AT LEAST one development instance, although best practice for Oracle Apps is to have (1) stand-alone patch database (2) development (3) test and (4) production. How you actually work depends on your software configuration control practices, and your tolerance for pain. Allowing developers to test in production or to add programs, change applications setting into production without complete testing is a receipe for disaster.
|
