mem.allocguestlargepage and mem.allocguestremotelargepage

[Tree_Dude]

What is the difference between these 2 settings? I am trying to maximize my memory utilization as we are adding VMs at a pace I did not anticipate and I have been denied a RAM upgrade till next year (I fear an HA event could be very bad). I read setting mem.allocguestlargepage to 0 will increase my memory sharing at the expense of some CPU (which is fine, we have lots). Should I touch mem.allocguestremotelargepage?

[admin · Accepted solution]

AllocGuestRemoteLargePage refers to 2MB page allocation on non-local NUMA nodes.  If you disable AllocGuestLargePage, remote 2MB page allocation will be disabled as well.  However, this should be unnecessary.  In response to memory pressure, large pages will eventually be splintered to take advantage of small page sharing opportunities.

[Tree_Dude]

I understand that. But I am trying to get a better gauge of what I truly have free rather than wait for the host memory to be in contention (btw I have no idea what the host uses to determine when TPS should kick in and the KBs I have read don't outline it clearly).

At the moment I don't mind the performance hit in return for better visibility on our memory utilization. Our hosts are brand new and have plenty of CPU to spare.