TPS has been around for a long time, and its purpose is to save memory at the host level. It is similar to storage deduplication, but this time focusing on the memory.
When multiple instances of VMs are run on the same ESXi hypervisor and access the same memory pages, they are stored only once. With TPS, the hypervisor will eliminate the redundant memory pages by mapping the identical content in only one memory page in the physical memory.
The TPS mechanism runs in the background and calculates a hash of the memory page. Those hashes are stored in a hash table and they are compared to each other by the ESXi server. If the ESXi kernel discovers two corresponding hashes, it will compare the content of the memory page. If the content is ...
