TPS: Transparent Page Sharing. What does it do ? Well think about all the “memory blocks” that are used by the VMs inside an ESXi Host RAM. These blocks “map” certain areas running inside VM RAM on the host.
What happens when you a certain number of VMs running a single kind of OS using the same “memory blocks” over and over again? Thats right, wastage of precious RAM, well used to precious a few years ago but not today, when my phone has more RAM than my first desktop.
So VMware have decided, after much deliberation I am sure, that TPS won’t be enabled by default in the next version of vSphere (vSphere 6.0). Its also made it very clear, that a number of patches will be released to ensure the same thing is done on the pre-existing systems that are in production today (ESXi 5.0+).
After applying the patch, will it affect the amount of RAM being utilised in the hosts? Yes. It will be. If you look at the hosts today, you will be able to see how much TPS has been helping in consumption of RAM or lack of it. You can see it by using esxtop on any ESXi Host or you can also use the new ‘fling’ Visual esxtop. Also TPS was not necessarily useful if you did enable Large Pages (which increases the block size to 2MB from 4kB).
Personally I have never seen more than 4-6GB worth of TPS saving on any given hosts, irrespective of what the host was being used for. And that 6GB Max was on a host running Citrix XenDesktop 5.6 VDI environment. We can purchase servers with almost 2TB of RAM installed today, which I am sure will keep increasing as the technology used to make those memory chips becomes smarter and smaller. I read somewhere that an ‘organic’ chip is being developed, which would have a capacity of 2000GB in the size of teardrop. So all in all the death of TPS was inevitable as memory chips started to shrink and security aspect of sharing pages was going to be a hot contender. You can read more about the ‘default disabling’ of TPS on VMware official KB Article at