Hi,
I am trying to understand Transparent Page Sharing and if I am reading this knowledge base article http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=102109... correctly, TPS will only kick in when memory starts to become overcommitted.
Is this true in real world situations that you guys are familiar with? We have a use case where we will be overcommitting memory and we think tps is one way we could do this because all of our vm's are the same exact image.
Thanks for any input.
TPS is working always with any license. But latest CPU generations offer big memory pages (2MB) which in theory should work faster than 4KB. However, in my environment there were no problems with CPU performance, while memory was always a limiting factor. There is a little chance that 2MB pages would be deduplicated unless they are filled with zeroes. When memory is overcommited and memory usage is close to 100% ESXi switches to 4KB pages and TPS starts to improve things.
You can switch to 4KB pages manually in Mem section of advanced settings (just disable big pages usage) to get TPS working all the time.
Maybe I'm missing it but how do I force the 4kb pages and disable big pages usage? I took a look at the advanced settings and the mem settings but don't see anything that correlates.
THanks for the help!
Try this setting.
Mem.AllocGuestLargePage = 0
And don't forget that you have to reboot the host after disabling large pages allocation.