Can someone explain why it is not possible to give high level recommendations for memory usage on Oracle running on Linux?
For example, if we have a 256GB RAM server and 1TB of accessed data in the Windows/SQL Server world, we could allocate >=200GB to SQL Server alone and the rest to the OS if we were only running one DB instance. Data is in the buffer cache and life makes sense as it should....In the Oracle world, there appears to no such recommendation. Supposedly if we did this in the Oracle world, performance might suffer. So let's say a VERY rough guesstimate is that for a Linux server with <=64GB RAM we could get away with not using Huge Pages and that for a Linux server with >64GB RAM we might want to consider using Huge Pages. So let's say we have a Linux server with 256GB RAM and we are using Huge Pages and the database is currently only allocated 20GB AMM and the actively accessed data size is 1TB. Why would we not set Oracle AMM to use most of the memory so that we keep most things in cache but we still leave enough for the OS? So set it to, say, 200GB RAM for Oracle and 56GB for the OS? Why would it ever be a BAD thing to increase memory as long as we had enough left for the OS/other programs? Oracle keeps track of memory much differently than Windows and it cannot access memory as fast? If so, is there a Oracle document that shows the internals as to why?