By TimRead on Aug 09, 2012
Even with just these four layers, the potential combinations expand very quickly: Two options for each layer leads to sixteen different stacks (24), three results in eighty-one (34) possibilities. This assuming that all options are available at each level, but does not take into consideration patching to the server firmware, the operating system, the database, or the middleware! So it's fairly clear that you must restrict yourself to one, or possibly two options at each layer to give yourself a chance of gaining, and maintaining, a grip on the complexity in your data center.
So what options do you pick? It's hard to be prescriptive here, but in general the latest versions of software have the most features, and fix the majority of the high priority bugs known on the previous release. Consequently, there is value in building on standard components that are as close to the leading edge as is reasonable. For example, you might pick Oracle Solaris 10 8/11 with a view to adopting Oracle Solaris 11 11/11 if and when the next release of Oracle Solaris 11 becomes available. In addition, you might choose to employ Oracle Grid Infrastructure 11g Release 2 as your primary availability platform for either 18.104.22.168 or 10.2.0.x Oracle databases.
Having said all that, there is still room for exception handling. Some applications just cannot be coerced onto one of your standard platforms, so they must be treated as exceptions. But again, you must keep the number of these down to keep complexity under control as you marshal your resources for the consolidation phase.