We have a SAP installation running the main server with 35GB ram and the app server running with 65GB ram. From the operational system perspective the main server has a fixed allocation of 31GB ram dedicated to one application, leaving 4GB ram for the operational system (Windows 2012 R2 64bits). From this remaining 4GB ram it happens often that we use almost 100%. Several times I log into the system and see 300MB ram free.
From the application perspective, we never have major errors on the main server, because it has 31GB ram for free usage and it uses usually 40-60% from that. The app server has some peaks but usually not bigger than 80% usage, so I don't believe we have problems there.
But what should be the adequate amount of free memory for the main server to be ran in the perspective of the Windows operational system?
When verifying best practices, I found generic information for critical systems that defines a conservative treshold of 33% of maximum usage. Whenever reaching that amount, the capacity should be reviewed. I would assume this is more a capacity planning question, but it is rather technical. If the system is maxing out the memory usage, should I necessarily raise the memory available? To which levels should I raise the memory in this scenario? Should I simply trust a dinamic setting from the virtual machine application manager? Thank you.
