Latest installment on Golden Gate licensing: I now have clarity on how the pricing works – the trouble is that, in its desire to remain consistent across all its products, Oracle may have made Golden Gate costs more difficult to justfy for individual projects.
The key to licensing Golden Gate is that you have to license all the CPU’s in both the source and target systems. As usual, the way you count “Processors” is to apply core factors and think about partioning (what are the “hard partitions” that the software occupies). This could add up to many tens of Procs in the large-scale, highly available systems that clients want to use Golden Gate with. Each ten Procs equals about £100k of licence cost at list price.
But here’s the issue: requirements for Golden Gate are in many cases time-bound. If you are migrating data around a database upgrade then the chances are the project will only need the Golden Gate licences for less than six months. When Golden Gate was a separate company the perpetual licences they provided were restricted to particular servers. That way if a large company wanted to migrate data in another application later they would need to buy new licences (to cover the different servers).
However, Oracle has now changed the licence model so that they are not restricted to particular servers – so as to fit in with the Oracle standard approach. Those licences are more flexible because they can be moved around a large organisation from project to project. The problem that does not fit with how projects are budgeted. The new Oracle model means that the Golden Gate licences for the first project are hugely expensive and the cost to subsequent projects is nil!
It remains to be seen whether this has an impact on Golden Gate sales. My guess is that it will elongate them – Oracle will need to find multiple project to fund a purchase or convince a central budget-holder to buy on behalf of many projects. Write a reply to tell me what you think.