Second Level Address Translation (SLAT) is a technology designed to reduce the overhead in a hypervisor CPU while easing the management of a hypervisor's memory and is often found with Hyper-V. These capabilities can improve the performance of virtual machines and ease the overhead of hypervisor performance on its host system.
SLAT works by adding a cache in the system's processor that will hold all of the recent page table mappings the hypervisor generated; this is called a Translation Lookaside Buffer (TLB). When there's a request for a virtual-to-physical address translation, the hypervisor first contacts the TLB for mapping information and, if there's a match, the hypervisor receives the physical address. This occurs without needing CPU over or memory to translate, hence the "second level" part of address translation.
If no match is found, the hypervisor will perform a normal address translation by checking its traditional page table, saving the translation in the TLB if it needs to be referenced in the future.